feat(server): add recipe constraints and update generation workspace (#2217)

.github/workflows/build.yml

@@ -26,6 +26,7 @@ on:
         "ggml",
         "examples/server/frontend",
         "examples/server/frontend/**",
+        "examples/server/recipes/*.json",
         "start-server.sh",
         "start-server.bat",
       ]
@@ -46,6 +47,7 @@ on:
         "ggml",
         "examples/server/frontend",
         "examples/server/frontend/**",
+        "examples/server/recipes/*.json",
         "start-server.sh",
         "start-server.bat",
       ]

docs/image_preprocessing.md

@@ -100,10 +100,13 @@ for individual references, identity images, or video control frames.
 --image-preprocess "target=control-frame,index=2,canny=true"
 ```
 
-Init and mask sources must have the same dimensions. The mask inherits the
-init crop, resize, and padding coordinates, while retaining its own filter,
-padding value, and Canny setting. Conflicting mask geometry is rejected. An
-omitted mask remains absent until the original pipeline creates its default mask.
+If init and mask source dimensions differ, the mask is first center-cropped to
+the init source aspect ratio and resized to the init source dimensions using
+nearest-exact interpolation. The mask then inherits the init crop, resize, and
+padding coordinates, while retaining its own filter, padding value, and Canny
+setting for that transform. Explicit mask geometry is evaluated after this
+alignment; conflicting geometry is rejected. An omitted mask remains absent
+until the original pipeline creates its default mask.
 
 ## Downstream behavior
 

examples/common/common.cpp

@@ -1959,6 +1959,7 @@ static bool parse_image_json_field(const json& parent,
     return decode_base64_image(parent.at(key).get<std::string>(), channels, 0, 0, out_image);
 }
 
+/** Parses a JSON array of base64 image strings into owned images. */
 static bool parse_image_array_json_field(const json& parent,
                                          const char* key,
                                          int channels,
@@ -1988,6 +1989,84 @@ static bool parse_image_array_json_field(const json& parent,
     return true;
 }
 
+/** Parses a JSON array of base64 WAV strings into owned audio clips. */
+static bool parse_audio_array_json_field(const json& parent,
+                                         const char* key,
+                                         std::vector<SDAudioOwner>& out_audios) {
+    if (!parent.contains(key)) {
+        return true;
+    }
+    if (parent.at(key).is_null()) {
+        out_audios.clear();
+        return true;
+    }
+    if (!parent.at(key).is_array()) {
+        return false;
+    }
+
+    out_audios.clear();
+    for (const auto& item : parent.at(key)) {
+        if (!item.is_string()) {
+            return false;
+        }
+        std::string encoded = item.get<std::string>();
+        auto comma_pos      = encoded.find(',');
+        if (comma_pos != std::string::npos) {
+            encoded = encoded.substr(comma_pos + 1);
+        }
+        std::vector<uint8_t> audio_bytes = decode_base64_bytes(encoded);
+        if (audio_bytes.empty()) {
+            return false;
+        }
+        std::vector<float> samples;
+        uint32_t sample_rate = 0;
+        uint32_t channels    = 0;
+        if (!load_wav_from_memory(audio_bytes.data(), audio_bytes.size(), samples, sample_rate, channels)) {
+            return false;
+        }
+        SDAudioOwner audio;
+        audio.reset(std::move(samples), sample_rate, channels);
+        out_audios.push_back(std::move(audio));
+    }
+    return true;
+}
+
+/** Parses a JSON array of per-video base64 frame arrays into owned videos. */
+static bool parse_video_array_json_field(const json& parent,
+                                         const char* key,
+                                         std::vector<std::vector<SDImageOwner>>& out_videos) {
+    if (!parent.contains(key)) {
+        return true;
+    }
+    if (parent.at(key).is_null()) {
+        out_videos.clear();
+        return true;
+    }
+    if (!parent.at(key).is_array()) {
+        return false;
+    }
+
+    out_videos.clear();
+    for (const auto& video : parent.at(key)) {
+        if (!video.is_array() || video.empty()) {
+            return false;
+        }
+        std::vector<SDImageOwner> frames;
+        for (const auto& frame : video) {
+            if (!frame.is_string()) {
+                return false;
+            }
+            SDImageOwner image;
+            if (!decode_base64_image(frame.get<std::string>(), 0, 0, 0, image)) {
+                return false;
+            }
+            frames.push_back(std::move(image));
+        }
+        out_videos.push_back(std::move(frames));
+    }
+    return true;
+}
+
 static bool parse_lora_json_field(const json& parent,
                                   const std::function<std::string(const std::string&)>& lora_path_resolver,
                                   std::map<std::string, float>& lora_map,
@@ -2102,6 +2181,7 @@ bool SDGenerationParams::parse_image_preprocess_json(const std::string& json_str
     return true;
 }
 
+/** Populates generation params from a JSON payload (server APIs); returns false on any malformed field. */
 bool SDGenerationParams::from_json_str(
     const std::string& json_str,
     const std::function<std::string(const std::string&)>& lora_path_resolver) {
@@ -2330,8 +2410,11 @@ bool SDGenerationParams::from_json_str(
         return false;
     }
     if (!parse_image_array_json_field(j, "ref_images", 0, ref_images) ||
-        !parse_image_array_json_field(j, "control_frames", 3, control_frames)) {
-        LOG_ERROR("invalid input image array");
+        !parse_image_array_json_field(j, "control_frames", 3, control_frames) ||
+        !parse_audio_array_json_field(j, "ref_audios", ref_audios) ||
+        !parse_audio_array_json_field(j, "ref_video_audios", ref_video_audios) ||
+        !parse_video_array_json_field(j, "ref_videos", ref_videos)) {
+        LOG_ERROR("invalid input image/audio/video array");
         return false;
     }
 
@@ -2583,6 +2666,7 @@ bool SDGenerationParams::resolve(const std::string& lora_model_dir, const std::s
     return true;
 }
 
+/** Validates params for a generation mode; enforces mode-specific input rules. */
 bool SDGenerationParams::validate(SDMode mode) {
     if (!image_preprocess.empty() && mode != IMG_GEN && mode != VID_GEN) {
         LOG_ERROR("--image-preprocess requires img_gen or vid_gen mode");
@@ -2636,12 +2720,13 @@ bool SDGenerationParams::validate(SDMode mode) {
         return false;
     }
 
-    if (ref_video_audio_paths.size() > ref_video_paths.size()) {
+    if (ref_video_audio_paths.size() + ref_video_audios.size() > ref_video_paths.size() + ref_videos.size()) {
         LOG_ERROR("error: each --ref-video-audio needs a corresponding --ref-video");
         return false;
     }
 
-    if (mode != VID_GEN && (!ref_video_paths.empty() || !ref_video_audio_paths.empty() || !ref_audio_paths.empty())) {
+    if (mode != VID_GEN && (!ref_video_paths.empty() || !ref_video_audio_paths.empty() || !ref_audio_paths.empty() ||
+                            !ref_videos.empty() || !ref_video_audios.empty() || !ref_audios.empty())) {
         LOG_ERROR("error: reference video and audio inputs require vid_gen mode");
         return false;
     }

examples/common/media_io.cpp

@@ -1437,70 +1437,94 @@ static uint32_t read_le32(const uint8_t* data) {
            (static_cast<uint32_t>(data[3]) << 24);
 }
 
+/** Loads a WAV file from disk and delegates decoding to load_wav_from_memory(). */
 bool load_wav_from_file(const std::string& path,
                         std::vector<float>& interleaved_samples,
                         uint32_t& sample_rate,
                         uint32_t& channels) {
+    std::ifstream file(fs::u8path(path), std::ios::binary);
+    if (!file) {
+        interleaved_samples.clear();
+        sample_rate = 0;
+        channels    = 0;
+        return false;
+    }
+    file.seekg(0, std::ios::end);
+    const std::streamoff file_size = file.tellg();
+    file.seekg(0, std::ios::beg);
+    if (file_size < 0) {
+        interleaved_samples.clear();
+        sample_rate = 0;
+        channels    = 0;
+        return false;
+    }
+    std::vector<uint8_t> buffer(static_cast<size_t>(file_size));
+    if (file_size > 0 && !file.read(reinterpret_cast<char*>(buffer.data()), file_size)) {
+        interleaved_samples.clear();
+        sample_rate = 0;
+        channels    = 0;
+        return false;
+    }
+    return load_wav_from_memory(buffer.data(), buffer.size(), interleaved_samples, sample_rate, channels);
+}
+
+/** Parses a RIFF/WAVE buffer in place; compares chunk sizes against the remaining
+ * bytes so payload-controlled sizes cannot wrap size_t on 32-bit builds. */
+bool load_wav_from_memory(const uint8_t* data,
+                          size_t size,
+                          std::vector<float>& interleaved_samples,
+                          uint32_t& sample_rate,
+                          uint32_t& channels) {
     interleaved_samples.clear();
     sample_rate = 0;
     channels    = 0;
 
-    std::ifstream file(fs::u8path(path), std::ios::binary);
-    uint8_t riff_header[12];
-    if (!file.read(reinterpret_cast<char*>(riff_header), sizeof(riff_header)) ||
-        std::memcmp(riff_header, "RIFF", 4) != 0 ||
-        std::memcmp(riff_header + 8, "WAVE", 4) != 0) {
+    if (data == nullptr ||
+        size < 12 ||
+        std::memcmp(data, "RIFF", 4) != 0 ||
+        std::memcmp(data + 8, "WAVE", 4) != 0) {
         return false;
     }
 
     uint16_t audio_format    = 0;
     uint16_t bits_per_sample = 0;
     uint16_t block_align     = 0;
-    std::streampos data_pos  = std::streampos(-1);
+    size_t data_pos          = SIZE_MAX;
     uint32_t data_size       = 0;
 
-    while (file.good()) {
-        uint8_t chunk_header[8];
-        if (!file.read(reinterpret_cast<char*>(chunk_header), sizeof(chunk_header))) {
-            break;
-        }
-        uint32_t chunk_size           = read_le32(chunk_header + 4);
-        std::streampos chunk_data_pos = file.tellg();
+    size_t pos = 12;
+    while (pos + 8 <= size) {
+        uint32_t chunk_size    = read_le32(data + pos + 4);
+        size_t chunk_data_pos  = pos + 8;
+        const size_t remaining = size - chunk_data_pos;  // safe: chunk_data_pos <= size
+        // Chunk sizes come from the payload, so compare against the remaining bytes
+        // instead of adding them to offsets (avoids size_t wrap on 32-bit builds).
+        const bool data_fits = static_cast<size_t>(chunk_size) <= remaining;
 
-        if (std::memcmp(chunk_header, "fmt ", 4) == 0) {
-            if (chunk_size < 16) {
+        if (std::memcmp(data + pos, "fmt ", 4) == 0) {
+            if (chunk_size < 16 || !data_fits) {
                 return false;
             }
-            std::vector<uint8_t> fmt(chunk_size);
-            if (!file.read(reinterpret_cast<char*>(fmt.data()), chunk_size)) {
-                return false;
-            }
-            audio_format    = read_le16(fmt.data());
-            channels        = read_le16(fmt.data() + 2);
-            sample_rate     = read_le32(fmt.data() + 4);
-            block_align     = read_le16(fmt.data() + 12);
-            bits_per_sample = read_le16(fmt.data() + 14);
+            audio_format    = read_le16(data + chunk_data_pos);
+            channels        = read_le16(data + chunk_data_pos + 2);
+            sample_rate     = read_le32(data + chunk_data_pos + 4);
+            block_align     = read_le16(data + chunk_data_pos + 12);
+            bits_per_sample = read_le16(data + chunk_data_pos + 14);
             if (audio_format == 0xfffe && chunk_size >= 40) {
-                audio_format = read_le16(fmt.data() + 24);
+                audio_format = read_le16(data + chunk_data_pos + 24);
             }
-        } else if (std::memcmp(chunk_header, "data", 4) == 0) {
+        } else if (std::memcmp(data + pos, "data", 4) == 0) {
             data_pos  = chunk_data_pos;
             data_size = chunk_size;
-            file.seekg(chunk_size, std::ios::cur);
-        } else {
-            file.seekg(chunk_size, std::ios::cur);
-        }
-
-        if (!file.good()) {
-            break;
-        }
-        if ((chunk_size & 1) != 0) {
-            file.seekg(1, std::ios::cur);
         }
+        // Only advance past a chunk whose size and pad byte fully fit (keeps the sum
+        // overflow-free); a chunk truncated at the buffer end terminates the walk.
+        const bool chunk_fits = data_fits && ((chunk_size & 1) == 0 || static_cast<size_t>(chunk_size) < remaining);
+        pos                   = chunk_fits ? chunk_data_pos + static_cast<size_t>(chunk_size) + (chunk_size & 1) : size;
     }
 
     const uint32_t bytes_per_sample = (bits_per_sample + 7) / 8;
-    if (data_pos == std::streampos(-1) || data_size == 0 || channels == 0 || sample_rate == 0 ||
+    if (data_pos == SIZE_MAX || data_size == 0 || channels == 0 || sample_rate == 0 ||
         block_align == 0 || bytes_per_sample == 0 || block_align < channels * bytes_per_sample ||
         (audio_format != 1 && audio_format != 3)) {
         return false;
@@ -1510,16 +1534,14 @@ bool load_wav_from_file(const std::string& path,
     if (frame_count == 0 || frame_count > SIZE_MAX / channels) {
         return false;
     }
-    std::vector<uint8_t> pcm(data_size);
-    file.clear();
-    file.seekg(data_pos);
-    if (!file.read(reinterpret_cast<char*>(pcm.data()), data_size)) {
+    if (static_cast<size_t>(data_size) > size - data_pos) {
         return false;
     }
+    const uint8_t* pcm = data + data_pos;
 
     interleaved_samples.resize(static_cast<size_t>(frame_count * channels));
     for (uint64_t frame = 0; frame < frame_count; ++frame) {
-        const uint8_t* frame_data = pcm.data() + frame * block_align;
+        const uint8_t* frame_data = pcm + frame * block_align;
         for (uint32_t channel = 0; channel < channels; ++channel) {
             const uint8_t* sample_data = frame_data + channel * bytes_per_sample;
             float sample               = 0.0f;

examples/common/media_io.h

@@ -126,4 +126,11 @@ bool load_wav_from_file(const std::string& path,
                         uint32_t& sample_rate,
                         uint32_t& channels);
 
+/** Decodes a WAV buffer (PCM 8/16/24/32 or IEEE float 32/64) into interleaved samples. */
+bool load_wav_from_memory(const uint8_t* data,
+                          size_t size,
+                          std::vector<float>& interleaved_samples,
+                          uint32_t& sample_rate,
+                          uint32_t& channels);
+
 #endif  // __MEDIA_IO_H__

examples/server/README.md

@@ -89,8 +89,8 @@ FLUX 1/2/3, Chroma, Qwen Image and editing variants; Z-Image, Anima, ERNIE,
 LENS, PiD and other supported image families; AnimateDiff, Wan, LTX,
 HunyuanVideo, LingBot Video and MiniMax-H3. Model-specific text encoder
 labels identify the expected encoder family and size. Loading a video model enables **Video Generation**; I2V also requires a
-start image. The Models panel can be collapsed and remembers its open state.
-Its header shows elapsed loading time from the status API and keeps the final
+start image. The model button below the prompt opens model and recipe selection.
+Its status summary shows elapsed loading time from the status API and keeps the final
 time after success or failure, including after a page refresh.
 Built-ins are embedded in the
 binary from [recipes/](recipes/), with one JSON file per recipe. Adding or

examples/server/api.md

@@ -183,6 +183,9 @@ Default-enabled features are `img2img`, `mask_image`, `lora`, `vae_tiling`,
 | `clip_skip` | CLIP Skip |
 | `img2img`, `mask_image` | Image initialization/strength and masks |
 | `ref_images` | Reference image inputs |
+| `ref_videos` | Reference videos as frame arrays |
+| `ref_audios` | Standalone reference audio; runtime requires an Audio VAE or Audio Encoder |
+| `ref_video_audios` | Soundtracks paired with reference videos; runtime requires an Audio VAE |
 | `controlnet` | Control image and ControlNet strength |
 | `control_image` | Control image without a separate ControlNet (FLUX Canny/Depth) |
 | `ip_adapter` | IP-Adapter image and strength |
@@ -241,6 +244,7 @@ the same one-recipe-per-file schema:
   "id": "my-flux",
   "name": "My FLUX",
   "description": "FLUX.1 dev",
+  "size_multiple": 16,
   "model_url": "https://huggingface.co/black-forest-labs/FLUX.1-dev",
   "slots": [
     {"id": "diffusion_model", "required": true},
@@ -256,7 +260,8 @@ the same one-recipe-per-file schema:
 
 IDs contain 1–128 ASCII letters, digits, dots, underscores or hyphens.
 `name`, `schema_version: 1` and `slots` are required; `description` and
-`model_url`, `required_any`, `defaults`, `context`, `features` and `disabled_features` are optional. Each slot requires only `id` and boolean `required`.
+`model_url`, `required_any`, `defaults`, `context`, `features`, `disabled_features`
+and `size_multiple`, `video_constraints` are optional. Each slot requires only `id` and boolean `required`.
 The server supplies the default label and search directories for each slot.
 An optional nonempty `label` overrides the display name, for example
 `{"id": "diffusion_model", "label": "Low-noise diffusion model", "required": true}`.
@@ -264,6 +269,45 @@ Legacy `category` fields remain accepted, but must match one of the slot's
 supported directories; they cannot redefine or restrict its search directories.
 At least one required `model` or `diffusion_model` slot is necessary.
 
+`size_multiple` is a positive integer from 1 to 4096 that declares the width and
+height increment used by the web UI. It defaults to `8` when omitted; the UI also
+uses `8` before a recipe is selected. For example, `"size_multiple": 16` makes
+generated size presets multiples of 16. Exact dimensions that do not match the
+increment require correction before submission. This is recipe metadata, not a
+generation parameter, and does not change the core model's alignment behavior.
+
+`video_constraints` is optional metadata for the web UI's Frames / Duration
+controls. It does not override the runtime's model-specific rules. For example:
+
+```json
+"video_constraints": {
+  "frame_step": 4,
+  "frame_offset": 1,
+  "frame_rounding": "floor",
+  "min_frames": 1
+}
+```
+
+Legal frame counts are `frame_step * n + frame_offset`, with integer `n >= 0`,
+within the inclusive `min_frames` / `max_frames` range. All fields are optional:
+
+| Field | Default | Accepted values |
+| --- | --- | --- |
+| `frame_step` | `1` | Integer 1–4096 |
+| `frame_offset` | `0` | Integer 0 to `frame_step - 1` |
+| `frame_rounding` | `"floor"` | `"floor"` or `"ceil"` |
+| `min_frames` | `1` | Integer 1–4096 |
+| `max_frames` | `4096` | Integer 1–4096, at least `min_frames` |
+| `fixed_fps` | Unset | Integer 1–240; locks the FPS control |
+
+The range must contain at least one legal frame count. Unknown keys are rejected.
+Omitting the object adds no frame alignment in the UI. `defaults.fps` only sets
+the initial FPS; it does not lock it. The UI multiplies requested duration by FPS,
+aligns using the declared rounding direction, then clamps to the legal range.
+Duration stays in the UI; generation requests still send only `video_frames` and
+`fps`. Switching input modes preserves the aligned length. Changing FPS keeps
+frames in Frames mode, and keeps the requested seconds in Duration mode.
+
 `model_url` is an optional HTTPS Hugging Face model repository link, shown beside
 the recipe description. Built-ins link to the model's author/publisher, not repacks.
 `required_any` contains groups of declared slot IDs. At least one model in each
@@ -1690,15 +1734,26 @@ When omitted, backend defaults apply to these fields:
 
 ### Frame Count Semantics
 
-`video_frames` is the requested target length, but the current core video path internally normalizes the effective frame count to the largest `4n + 1` value that does not exceed the requested count.
+`video_frames` is the requested target length. The core normalizes it according
+to the loaded model:
 
-Examples:
+| Model family | Frame alignment | Fixed FPS |
+| --- | --- | --- |
+| Wan, HunyuanVideo, LingBot | `4n + 1`, round down | Wan 2.2 S2V: 16 |
+| LTX | `8n + 1`, round down | None |
+| MiniMax-H3 | `17n + 5`, round up, minimum 5 | 24 |
+| AnimateDiff | Whole frames, maximum 32 | None |
+
+Wan examples:
 
 - `video_frames = 33` stays `33`
 - `video_frames = 34` becomes `33`
 - `video_frames = 32` becomes `29`
 
-The completed job payload includes the actual decoded `frame_count`.
+At 16 FPS, a 5-second target requests 80 frames and aligns to 77 frames for Wan,
+or approximately 4.813 seconds. Audio-driven generation and temporal upscaling
+can change the final length. Use the completed job's actual decoded `frame_count`
+and `fps` for the final playback duration.
 
 ### Completion Result
 

examples/server/model_catalog.cpp

@@ -291,6 +291,7 @@ static void apply_recipe_context(const json& settings, SDContextParams& params)
     }
 }
 
+// Known recipe feature names with their defaults; recipes may flip any of these.
 static const std::map<std::string, bool> k_recipe_features = {
     {"img2img", true},
     {"mask_image", true},
@@ -312,6 +313,9 @@ static const std::map<std::string, bool> k_recipe_features = {
     {"init_image", false},
     {"end_image", false},
     {"control_frames", false},
+    {"ref_videos", false},
+    {"ref_audios", false},
+    {"ref_video_audios", false},
     {"vace", false},
     {"high_noise", false},
     {"cache_unet", false},
@@ -365,17 +369,44 @@ json model_generation_features(const ServerRuntime& runtime) {
     if (runtime.models->active_recipe.value("source", "") == "command_line")
         return features;
     // Optional components only become usable once selected and loaded.
-    features["ip_adapter"]  = features.value("ip_adapter", false) && !runtime.ctx_params->ip_adapter_path.empty();
-    features["controlnet"]  = features.value("controlnet", false) && !runtime.ctx_params->control_net_path.empty();
-    features["high_noise"]  = features.value("high_noise", false) && !runtime.ctx_params->high_noise_diffusion_model_path.empty();
-    features["preview_tae"] = features.value("preview_tae", false) && !runtime.ctx_params->taesd_path.empty();
+    features["ip_adapter"]       = features.value("ip_adapter", false) && !runtime.ctx_params->ip_adapter_path.empty();
+    features["controlnet"]       = features.value("controlnet", false) && !runtime.ctx_params->control_net_path.empty();
+    features["high_noise"]       = features.value("high_noise", false) && !runtime.ctx_params->high_noise_diffusion_model_path.empty();
+    features["preview_tae"]      = features.value("preview_tae", false) && !runtime.ctx_params->taesd_path.empty();
+    features["ref_audios"]       = features.value("ref_audios", false) &&
+                                   (!runtime.ctx_params->audio_vae_path.empty() || !runtime.ctx_params->audio_encoder_path.empty());
+    features["ref_video_audios"] = features.value("ref_video_audios", false) && !runtime.ctx_params->audio_vae_path.empty();
     if (!runtime.ctx_params->pulid_weights_path.empty())
         features["slg"] = false;
     return features;
 }
 
+static void validate_video_constraints(const json& constraints) {
+    check_keys(constraints, {"frame_step", "frame_offset", "frame_rounding", "min_frames", "max_frames", "fixed_fps"});
+    for (const auto* key : {"frame_step", "frame_offset", "min_frames", "max_frames", "fixed_fps"}) {
+        const int minimum = std::string(key) == "frame_offset" ? 0 : 1;
+        const int maximum = std::string(key) == "fixed_fps" ? 240 : 4096;
+        if (constraints.contains(key) &&
+            (!constraints[key].is_number_integer() || constraints[key] < minimum || constraints[key] > maximum)) {
+            throw std::runtime_error(std::string("video_constraints.") + key + " is outside its integer range");
+        }
+    }
+    const int step    = constraints.value("frame_step", 1);
+    const int offset  = constraints.value("frame_offset", 0);
+    const int minimum = constraints.value("min_frames", 1);
+    const int maximum = constraints.value("max_frames", 4096);
+    if (offset >= step || minimum > maximum ||
+        offset + std::max(0, (minimum - offset + step - 1) / step) * step > maximum) {
+        throw std::runtime_error("video_constraints must describe a nonempty frame range with frame_offset < frame_step");
+    }
+    if (constraints.contains("frame_rounding") &&
+        constraints["frame_rounding"] != "floor" && constraints["frame_rounding"] != "ceil") {
+        throw std::runtime_error("video_constraints.frame_rounding must be floor or ceil");
+    }
+}
+
 static void validate_recipe(const json& recipe) {
-    check_keys(recipe, {"schema_version", "id", "name", "description", "model_url", "slots", "required_any", "defaults", "context", "features", "disabled_features"});
+    check_keys(recipe, {"schema_version", "id", "name", "description", "model_url", "slots", "required_any", "defaults", "context", "features", "disabled_features", "size_multiple", "video_constraints"});
     if (!recipe.at("schema_version").is_number_integer() || recipe.at("schema_version") != 1 ||
         !valid_identifier(recipe.at("id").get<std::string>()) || recipe.at("name").get<std::string>().empty()) {
         throw std::runtime_error("invalid schema_version, id or name");
@@ -383,6 +414,13 @@ static void validate_recipe(const json& recipe) {
     if (recipe.contains("description") && !recipe["description"].is_string()) {
         throw std::runtime_error("description must be a string");
     }
+    if (recipe.contains("size_multiple") &&
+        (!recipe["size_multiple"].is_number_integer() || recipe["size_multiple"] < 1 || recipe["size_multiple"] > 4096)) {
+        throw std::runtime_error("size_multiple must be an integer from 1 to 4096");
+    }
+    if (recipe.contains("video_constraints")) {
+        validate_video_constraints(recipe["video_constraints"]);
+    }
     if (recipe.contains("model_url")) {
         const auto url           = recipe.at("model_url").get<std::string>();
         const std::string prefix = "https://huggingface.co/";
@@ -435,6 +473,7 @@ static void validate_recipe(const json& recipe) {
 
 static json normalize_recipe(json recipe) {
     validate_recipe(recipe);
+    recipe["size_multiple"]      = recipe.value("size_multiple", 8);
     recipe["effective_features"] = resolve_recipe_features(recipe);
     for (auto& slot : recipe.at("slots")) {
         const auto& spec   = *find_model_slot(slot.at("id").get<std::string>());
@@ -598,6 +637,7 @@ void initialize_server_models(ServerRuntime& runtime) {
             {"description", "Uses the loaded command-line model and its generation defaults."},
             {"slots", json::array()},
             {"defaults", json::object()},
+            {"size_multiple", 8},
             {"features", features},
         };
         models.active_recipe["effective_features"] = resolve_recipe_features(models.active_recipe);

examples/server/recipes/README.md

@@ -77,3 +77,35 @@ These are built-in catalog rules. Users may keep personal presets in
 `<model-dir>/recipes/`; they need not add those presets to the built-in list.
 See [the recipe schema and feature reference](../api.md#user-recipe-format)
 for supported fields.
+
+## Output dimension alignment
+
+Set `size_multiple` when the runtime's output dimension alignment differs from
+the default of `8`. Match `GenerationRequest::align_image_size` in
+[`src/pipeline/request.cpp`](../../../src/pipeline/request.cpp): the VAE scale
+factor multiplied by the model definition's `diffusion_down_factor`. The scale
+comes from [`VAE::get_scale_factor`](../../../src/model/vae/vae.hpp), with the
+PiD override in [`diffusion_engine.cpp`](../../../src/pipeline/diffusion_engine.cpp);
+model geometry is defined under [`src/model/integrations`](../../../src/model/integrations/).
+
+Use the runtime's output alignment, not the nominal patch size: some models pad
+patches internally and crop their output back to the requested size. Pixel-space
+models with an alignment of `1` explicitly use `size_multiple: 1`. MiniT2I also
+has this alignment, but its separate fixed `512 × 512` output restriction still
+applies. Reference-image preprocessing constraints are independent of this field.
+
+## Video length constraints
+
+Video recipes declare `video_constraints` to match `align_video_frames` in
+[`diffusion_engine.cpp`](../../../src/pipeline/diffusion_engine.cpp) and the FPS
+and AnimateDiff overrides in
+[`request.cpp`](../../../src/pipeline/request.cpp). Wan, HunyuanVideo and LingBot
+round down to `4n + 1`; LTX rounds down to `8n + 1`. MiniMax-H3 rounds up to
+`17n + 5`, with at least 5 frames and fixed 24 FPS. AnimateDiff accepts whole
+frames up to 32. Wan 2.2 S2V fixes FPS at 16.
+
+Declare fixed FPS only where the runtime enforces it; recommended FPS belongs
+in `defaults.fps`. These constraints guide UI conversion and estimates, not
+runtime behavior. Audio-driven output and temporal upscaling may still change
+the final frame count. See the [recipe schema](../api.md#user-recipe-format)
+for field defaults and validation rules.

examples/server/recipes/animatediff-sd15.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "animatediff-sd15",
   "name": "AnimateDiff SD 1.5",
+  "size_multiple": 64,
   "description": "Use Video Generation with an SD 1.5 checkpoint and compatible motion module.",
   "model_url": "https://huggingface.co/guoyww/animatediff",
   "slots": [
@@ -22,6 +23,13 @@
       "required": false
     }
   ],
+  "video_constraints": {
+    "frame_step": 1,
+    "frame_offset": 0,
+    "frame_rounding": "floor",
+    "min_frames": 1,
+    "max_frames": 32
+  },
   "defaults": {
     "sampling_method": "euler",
     "video_frames": 8,

examples/server/recipes/chroma-radiance.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "chroma-radiance",
   "name": "Chroma Radiance",
+  "size_multiple": 1,
   "description": "Pixel-space generation without a VAE. Also accepts the x0 and experimental patch-32 variants.",
   "model_url": "https://huggingface.co/lodestones/Chroma1-Radiance",
   "slots": [

examples/server/recipes/cosxl-edit.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "cosxl-edit",
   "name": "CosXL Edit",
+  "size_multiple": 64,
   "description": "Supply the image to edit as a reference image.",
   "model_url": "https://huggingface.co/stabilityai/cosxl",
   "slots": [

examples/server/recipes/defaults.md

@@ -5,6 +5,9 @@ Reviewed against model-publisher cards and reference implementations on 2026-10-
 These are starting values, not limits on supported resolutions or generation settings.
 Built-in step defaults above 20 are reduced to 20; the cited sources may recommend more steps.
 Users can still choose a higher step count.
+Except for PiD recipes, built-in dimensions above 1,048,576 pixels (1024 x 1024) are reduced to 1024 x 1024, even when the cited source uses a larger size.
+PiD recipes retain their 2048 x 2048 defaults.
+Users can still choose a higher resolution.
 When several aspect ratios or ranges are published, the selected example or square bucket is noted below.
 A dash means no recipe override: the available primary sources did not establish a fixed value, or the size depends on the input image.
 Ordinary DiT recipes explicitly use Euler; PiD retains its LCM sampling path.
@@ -27,7 +30,7 @@ For community edit adapters and other optional extensions, these are the base-mo
 | `flux1-canny` | 1 | 20 | 1024 × 1024 | [Source](https://huggingface.co/black-forest-labs/FLUX.1-Canny-dev)  |
 | `flux1-depth` | 1 | 20 | 1024 × 1024 | [Source](https://huggingface.co/black-forest-labs/FLUX.1-Depth-dev)  |
 | `flux1-dev` | 1 | 20 | 1024 × 1024 | [Source](https://huggingface.co/black-forest-labs/FLUX.1-dev)  |
-| `flux1-fill` | 1 | 20 | 1232 × 1632 | [Source](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev)  |
+| `flux1-fill` | 1 | 20 | 1024 × 1024 | [Source](https://huggingface.co/black-forest-labs/FLUX.1-Fill-dev)  |
 | `flux1-kontext` | 1 | 20 | — | [Source](https://raw.githubusercontent.com/huggingface/diffusers/main/src/diffusers/pipelines/flux/pipeline_flux_kontext.py) Output size normally follows the reference image; no fixed size is imposed by this recipe. |
 | `flux1-pulid` | 1 | 20 | 1024 × 1024 | [Source](https://huggingface.co/black-forest-labs/FLUX.1-dev) FLUX.1-dev defaults; identity conditioning is provided by PuLID. |
 | `flux1-schnell` | 1 | 4 | 1024 × 1024 | [Source](https://huggingface.co/black-forest-labs/FLUX.1-schnell)  |
@@ -38,7 +41,7 @@ For community edit adapters and other optional extensions, these are the base-mo
 | `flux2-klein-base-9b` | 4 | 20 | 1024 × 1024 | [Source](https://huggingface.co/black-forest-labs/FLUX.2-klein-base-9B)  |
 | `flux3-klein` | 1 | 16 | 1024 × 1024 | [Source](https://raw.githubusercontent.com/black-forest-labs/flux-3-image-klein/main/README.md)  |
 | `flux3-klein-base` | 4 | 20 | 1024 × 1024 | [Source](https://raw.githubusercontent.com/black-forest-labs/flux-3-image-klein/main/src/flux3/pipeline.py)  |
-| `hidream-o1-dev` | 1 | 20 | 2048 × 2048 | [Source](https://huggingface.co/HiDream-ai/HiDream-O1-Image-Dev)  |
+| `hidream-o1-dev` | 1 | 20 | 1024 × 1024 | [Source](https://huggingface.co/HiDream-ai/HiDream-O1-Image-Dev)  |
 | `hunyuan-video-15-i2v` | 6 | 20 | 1280 × 720 | [Source](https://huggingface.co/tencent/HunyuanVideo-1.5) 720p base I2V defaults; CFG/step-distilled and 480p checkpoints use different settings. |
 | `hunyuan-video-15-t2v` | 6 | 20 | 1280 × 720 | [Source](https://huggingface.co/tencent/HunyuanVideo-1.5) 720p base T2V defaults; CFG-distilled / 480p checkpoints need their published settings. |
 | `ideogram4` | 7 | 20 | 1024 × 1024 | [Source](https://raw.githubusercontent.com/ideogram-oss/ideogram4/main/docs/inference.md) Constant-guidance pipeline defaults; this recipe does not reproduce the CLI’s per-step guidance schedule. |
@@ -46,10 +49,10 @@ For community edit adapters and other optional extensions, these are the base-mo
 | `iris-3b` | 3 | 20 | 1024 × 1024 | [Source](https://huggingface.co/speridlabs/iris-3b)  |
 | `krea2-raw` | 3.5 | 20 | 1024 × 1024 | [Source](https://huggingface.co/krea/Krea-2-Raw) Base model defaults; edit requires a matching community LoRA with its own settings. |
 | `krea2-raw-edit` | 3.5 | 20 | 1024 × 1024 | [Source](https://huggingface.co/krea/Krea-2-Raw) Base model defaults; edit requires a matching community LoRA with its own settings. |
-| `krea2-turbo` | 1 | 8 | 2048 × 2048 | [Source](https://huggingface.co/krea/Krea-2-Turbo) Official CLI example; edit requires a matching community LoRA. |
-| `krea2-turbo-edit` | 1 | 8 | 2048 × 2048 | [Source](https://huggingface.co/krea/Krea-2-Turbo) Official CLI example; edit requires a matching community LoRA. |
-| `lens` | 5 | 20 | 1440 × 1440 | [Source](https://huggingface.co/microsoft/Lens) Official model card: RL model, default base_resolution 1440. |
-| `lens-turbo` | 1 | 4 | 1440 × 1440 | [Source](https://huggingface.co/microsoft/Lens-Turbo) Official model card: 4-step distilled model, default base_resolution 1440. |
+| `krea2-turbo` | 1 | 8 | 1024 × 1024 | [Source](https://huggingface.co/krea/Krea-2-Turbo) Official CLI example; edit requires a matching community LoRA. |
+| `krea2-turbo-edit` | 1 | 8 | 1024 × 1024 | [Source](https://huggingface.co/krea/Krea-2-Turbo) Official CLI example; edit requires a matching community LoRA. |
+| `lens` | 5 | 20 | 1024 × 1024 | [Source](https://huggingface.co/microsoft/Lens) Official model card: RL model, default base_resolution 1440. |
+| `lens-turbo` | 1 | 4 | 1024 × 1024 | [Source](https://huggingface.co/microsoft/Lens-Turbo) Official model card: 4-step distilled model, default base_resolution 1440. |
 | `lingbot-video-dense` | 3 | 20 | 832 × 480 | [Source](https://huggingface.co/robbyant/lingbot-video-dense-1.3b) Official first-stage inference example; optional refiner is separate. |
 | `lingbot-video-moe` | 3 | 20 | 832 × 480 | [Source](https://huggingface.co/robbyant/lingbot-video-moe-30b-a3b) Official inference example. |
 | `llada-image` | 5 | 20 | 1024 × 1024 | [Source](https://huggingface.co/inclusionAI/LLaDA-Image)  |
@@ -62,7 +65,7 @@ For community edit adapters and other optional extensions, these are the base-mo
 | `mage-flow` | 5 | 20 | 1024 × 1024 | [Source](https://huggingface.co/microsoft/Mage-Flow)  |
 | `mage-flow-edit` | 5 | 20 | 1024 × 1024 | [Source](https://huggingface.co/microsoft/Mage-Flow)  |
 | `mage-flow-turbo` | 1 | 4 | 1024 × 1024 | [Source](https://huggingface.co/microsoft/Mage-Flow)  |
-| `ming-image` | 1 | 12 | 2048 × 2048 | [Source](https://huggingface.co/inclusionAI/Ming-Image-0.1-Design)  |
+| `ming-image` | 1 | 12 | 1024 × 1024 | [Source](https://huggingface.co/inclusionAI/Ming-Image-0.1-Design)  |
 | `minimax-h3-fl2va` | 1 | — | 1344 × 768 | [Source](https://huggingface.co/docs/diffusers/main/en/api/pipelines/minimax_h3) Guidance-distilled; official 768p canvas. No fixed step count is prescribed here; the upstream API counts sigma grid points rather than sd.cpp denoising steps. |
 | `minimax-h3-ref2va` | 1 | — | 1344 × 768 | [Source](https://huggingface.co/docs/diffusers/main/en/api/pipelines/minimax_h3) Guidance-distilled; official 768p canvas. No fixed step count is prescribed here; the upstream API counts sigma grid points rather than sd.cpp denoising steps. |
 | `minit2i` | 2.5 | 20 | 512 × 512 | [Source](https://huggingface.co/MiniT2I/MiniT2I/raw/main/pipeline.py) B/16 example and native model size; L/16 uses CFG 6. |
@@ -81,9 +84,9 @@ For community edit adapters and other optional extensions, these are the base-mo
 | `pixart-alpha-1024` | 4.5 | 20 | 1024 × 1024 | [Source](https://raw.githubusercontent.com/huggingface/diffusers/main/src/diffusers/pipelines/pixart_alpha/pipeline_pixart_alpha.py)  |
 | `pixart-alpha-512` | 4.5 | 20 | 512 × 512 | [Source](https://raw.githubusercontent.com/huggingface/diffusers/main/src/diffusers/pipelines/pixart_alpha/pipeline_pixart_alpha.py)  |
 | `pixart-sigma` | 4.5 | 20 | 1024 × 1024 | [Source](https://raw.githubusercontent.com/huggingface/diffusers/main/src/diffusers/pipelines/pixart_alpha/pipeline_pixart_sigma.py)  |
-| `qwen-image` | 4 | 20 | 1328 × 1328 | [Source](https://huggingface.co/Qwen/Qwen-Image) Official square resolution bucket; true_cfg_scale maps to cfg_scale. |
-| `qwen-image-21` | 1 | 20 | 2048 × 2048 | [Source](https://raw.githubusercontent.com/QwenLM/Qwen-Image-2.1/main/README.md) Original checkpoint defaults; select the matching TAEQI2.1 to replace its VAE. |
-| `qwen-image-21-turbo` | 1 | 8 | 2048 × 2048 | [Model card](https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo), [fixed sigmas](https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo/blob/main/model_index.json). Standalone Turbo weights; the fixed 8-step schedule is applied on model load. |
+| `qwen-image` | 4 | 20 | 1024 × 1024 | [Source](https://huggingface.co/Qwen/Qwen-Image) The source uses a larger square resolution bucket; true_cfg_scale maps to cfg_scale. |
+| `qwen-image-21` | 1 | 20 | 1024 × 1024 | [Source](https://raw.githubusercontent.com/QwenLM/Qwen-Image-2.1/main/README.md) Original checkpoint sampling defaults; select the matching TAEQI2.1 to replace its VAE. |
+| `qwen-image-21-turbo` | 1 | 8 | 1024 × 1024 | [Model card](https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo), [fixed sigmas](https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo/blob/main/model_index.json). Standalone Turbo weights; the fixed 8-step schedule is applied on model load. |
 | `qwen-image-edit` | 4 | 20 | — | [Source](https://huggingface.co/Qwen/Qwen-Image-Edit) Official editing example; output dimensions follow reference-image aspect ratio. |
 | `qwen-image-edit-2509` | 4 | 20 | — | [Source](https://huggingface.co/Qwen/Qwen-Image-Edit-2509) Official editing example; output dimensions follow reference-image aspect ratio. |
 | `qwen-image-edit-2511` | 4 | 20 | — | [Source](https://huggingface.co/Qwen/Qwen-Image-Edit-2511) Official editing example; output dimensions follow reference-image aspect ratio. |
@@ -107,7 +110,7 @@ For community edit adapters and other optional extensions, these are the base-mo
 | `sdxl-turbo` | 1 | 1 | 512 × 512 | [Source](https://huggingface.co/stabilityai/sdxl-turbo)  |
 | `sdxs` | 1 | 1 | 512 × 512 | [Source](https://huggingface.co/IDKiro/sdxs-512-dreamshaper)  |
 | `sefi-image` | 4 | 20 | — | [Source](https://huggingface.co/SeFi-Image/SeFi-Image-5B-Base) Base/RL family defaults; Turbo checkpoints require 4 steps / CFG 1. |
-| `sensenova-u15` | 4 | 20 | 2048 × 2048 | [Source](https://raw.githubusercontent.com/OpenSenseNova/SenseNova-U1/main/examples/README.md)  |
+| `sensenova-u15` | 4 | 20 | 1024 × 1024 | [Source](https://raw.githubusercontent.com/OpenSenseNova/SenseNova-U1/main/examples/README.md)  |
 | `wan21-flf2v` | 5 | 20 | 1280 × 720 | [Source](https://raw.githubusercontent.com/Wan-Video/Wan2.1/main/generate.py)  |
 | `wan21-i2v` | 5 | 20 | 832 × 480 | [Source](https://raw.githubusercontent.com/Wan-Video/Wan2.1/main/generate.py) 480p base defaults; also accepts compatible SkyReels V2 and single Wan 2.2 I2V experts. |
 | `wan21-t2v` | 5 | 20 | 1280 × 720 | [Source](https://raw.githubusercontent.com/Wan-Video/Wan2.1/main/generate.py) 14B base defaults; single Wan 2.2 experts and acceleration LoRAs need matching settings. |

examples/server/recipes/ernie-image-turbo.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "ernie-image-turbo",
   "name": "ERNIE Image Turbo",
+  "size_multiple": 16,
   "model_url": "https://huggingface.co/baidu/ERNIE-Image-Turbo",
   "slots": [
     {

examples/server/recipes/ernie-image.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "ernie-image",
   "name": "ERNIE Image",
+  "size_multiple": 16,
   "model_url": "https://huggingface.co/baidu/ERNIE-Image",
   "slots": [
     {

examples/server/recipes/fibo2-dev.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "fibo2-dev",
   "name": "Fibo 2.0 Dev",
+  "size_multiple": 16,
   "description": "Text-to-image generation and native reference-image editing. Select a separate vision projector for editing when it is not bundled in the text encoder.",
   "model_url": "https://huggingface.co/briaai/Fibo-2.0-dev",
   "slots": [

examples/server/recipes/fibo2-edit-turbo.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "fibo2-edit-turbo",
   "name": "Fibo 2.0 Edit Turbo",
+  "size_multiple": 16,
   "description": "Dedicated reference-image editing checkpoint; supply reference images. Select a separate vision projector when it is not bundled in the text encoder.",
   "model_url": "https://huggingface.co/briaai/Fibo-2.0-edit-turbo",
   "slots": [

examples/server/recipes/flux1-fill.json

@@ -39,8 +39,8 @@
     "scheduler": "flux",
     "cfg_scale": 1,
     "steps": 20,
-    "width": 1232,
-    "height": 1632
+    "width": 1024,
+    "height": 1024
   },
   "features": [
     "clip_skip",

examples/server/recipes/flux2-dev.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "flux2-dev",
   "name": "FLUX.2 dev",
+  "size_multiple": 16,
   "description": "Supports text-to-image and reference-image editing.",
   "model_url": "https://huggingface.co/black-forest-labs/FLUX.2-dev",
   "slots": [

examples/server/recipes/flux2-klein-4b.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "flux2-klein-4b",
   "name": "FLUX.2 Klein 4B",
+  "size_multiple": 16,
   "description": "Supports reference-image editing. The FLUX.2 small decoder is an alternative VAE.",
   "model_url": "https://huggingface.co/black-forest-labs/FLUX.2-klein-4B",
   "slots": [

examples/server/recipes/flux2-klein-9b.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "flux2-klein-9b",
   "name": "FLUX.2 Klein 9B",
+  "size_multiple": 16,
   "description": "Supports text-to-image and reference-image editing.",
   "model_url": "https://huggingface.co/black-forest-labs/FLUX.2-klein-9B",
   "slots": [

examples/server/recipes/flux2-klein-base-4b.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "flux2-klein-base-4b",
   "name": "FLUX.2 Klein Base 4B",
+  "size_multiple": 16,
   "description": "Supports text-to-image and reference-image editing.",
   "model_url": "https://huggingface.co/black-forest-labs/FLUX.2-klein-base-4B",
   "slots": [

examples/server/recipes/flux2-klein-base-9b.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "flux2-klein-base-9b",
   "name": "FLUX.2 Klein Base 9B",
+  "size_multiple": 16,
   "description": "Use the 9B base transformer with Qwen 3 8B. Supports reference-image editing.",
   "model_url": "https://huggingface.co/black-forest-labs/FLUX.2-klein-base-9B",
   "slots": [

examples/server/recipes/flux3-klein-base.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "flux3-klein-base",
   "name": "FLUX.3 Image Klein Base",
+  "size_multiple": 16,
   "model_url": "https://huggingface.co/black-forest-labs/flux-3-image-klein-base",
   "slots": [
     {

examples/server/recipes/flux3-klein.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "flux3-klein",
   "name": "FLUX.3 Image Klein",
+  "size_multiple": 16,
   "description": "Supports text-to-image and reference-image editing.",
   "model_url": "https://huggingface.co/black-forest-labs/flux-3-image-klein",
   "slots": [

examples/server/recipes/hidream-o1-dev.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "hidream-o1-dev",
   "name": "HiDream O1 Image dev",
+  "size_multiple": 1,
   "description": "Bundled checkpoint; also supports reference-image editing.",
   "model_url": "https://huggingface.co/HiDream-ai/HiDream-O1-Image-Dev",
   "slots": [
@@ -13,8 +14,8 @@
   "defaults": {
     "cfg_scale": 1,
     "steps": 20,
-    "width": 2048,
-    "height": 2048,
+    "width": 1024,
+    "height": 1024,
     "sampling_method": "euler"
   },
   "features": [

examples/server/recipes/hunyuan-video-15-i2v.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "hunyuan-video-15-i2v",
   "name": "HunyuanVideo 1.5 I2V",
+  "size_multiple": 16,
   "description": "Select HunyuanVideo 1.5 I2V weights and supply a start image in Video Generation.",
   "model_url": "https://huggingface.co/tencent/HunyuanVideo-1.5",
   "slots": [
@@ -36,6 +37,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 4,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "sampling_method": "euler",
     "video_frames": 33,

examples/server/recipes/hunyuan-video-15-t2v.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "hunyuan-video-15-t2v",
   "name": "HunyuanVideo 1.5 T2V",
+  "size_multiple": 16,
   "description": "Select HunyuanVideo 1.5 T2V weights and use Video Generation. Use the separate I2V recipe for I2V weights.",
   "model_url": "https://huggingface.co/tencent/HunyuanVideo-1.5",
   "slots": [
@@ -36,6 +37,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 4,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "sampling_method": "euler",
     "video_frames": 33,

examples/server/recipes/ideogram4.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "ideogram4",
   "name": "Ideogram 4",
+  "size_multiple": 16,
   "description": "Select the matching unconditional transformer for classifier-free guidance.",
   "model_url": "https://huggingface.co/ideogram-ai/ideogram-4-fp8",
   "slots": [

examples/server/recipes/instruct-pix2pix.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "instruct-pix2pix",
   "name": "Instruct-Pix2Pix",
+  "size_multiple": 64,
   "description": "Supply the image to edit as a reference image.",
   "model_url": "https://huggingface.co/timbrooks/instruct-pix2pix",
   "slots": [

examples/server/recipes/iris-3b.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "iris-3b",
   "name": "Iris 3B",
+  "size_multiple": 16,
   "description": "Pixel-space model; no VAE is required.",
   "model_url": "https://huggingface.co/speridlabs/iris-3b",
   "slots": [

examples/server/recipes/krea2-turbo-edit.json

@@ -39,8 +39,8 @@
     "sampling_method": "euler",
     "cfg_scale": 1,
     "steps": 8,
-    "width": 2048,
-    "height": 2048
+    "width": 1024,
+    "height": 1024
   },
   "features": [
     "flow_shift",

examples/server/recipes/krea2-turbo.json

@@ -32,8 +32,8 @@
   "defaults": {
     "cfg_scale": 1,
     "steps": 8,
-    "width": 2048,
-    "height": 2048,
+    "width": 1024,
+    "height": 1024,
     "sampling_method": "euler"
   },
   "features": [

examples/server/recipes/lens-turbo.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "lens-turbo",
   "name": "LENS Turbo",
+  "size_multiple": 16,
   "model_url": "https://huggingface.co/microsoft/Lens-Turbo",
   "slots": [
     {
@@ -36,8 +37,8 @@
   "defaults": {
     "cfg_scale": 1,
     "steps": 4,
-    "width": 1440,
-    "height": 1440,
+    "width": 1024,
+    "height": 1024,
     "sampling_method": "euler"
   },
   "features": [

examples/server/recipes/lens.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "lens",
   "name": "LENS",
+  "size_multiple": 16,
   "model_url": "https://huggingface.co/microsoft/Lens",
   "slots": [
     {
@@ -36,8 +37,8 @@
   "defaults": {
     "cfg_scale": 5,
     "steps": 20,
-    "width": 1440,
-    "height": 1440,
+    "width": 1024,
+    "height": 1024,
     "sampling_method": "euler"
   },
   "features": [

examples/server/recipes/lingbot-video-dense.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "lingbot-video-dense",
   "name": "LingBot Video Dense 1.3B",
+  "size_multiple": 16,
   "description": "Use Video Generation.",
   "model_url": "https://huggingface.co/robbyant/lingbot-video-dense-1.3b",
   "slots": [
@@ -30,6 +31,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 4,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "video_frames": 33,
     "cfg_scale": 3,

examples/server/recipes/lingbot-video-moe.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "lingbot-video-moe",
   "name": "LingBot Video MoE 30B A3B",
+  "size_multiple": 16,
   "description": "Use Video Generation.",
   "model_url": "https://huggingface.co/robbyant/lingbot-video-moe-30b-a3b",
   "slots": [
@@ -30,6 +31,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 4,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "video_frames": 33,
     "cfg_scale": 3,

examples/server/recipes/llada-image-turbo.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "llada-image-turbo",
   "name": "LLaDA Image Turbo",
+  "size_multiple": 16,
   "description": "Use matching turbo text encoder and connectors. Editing needs connectors that include the SigVQ encoder.",
   "model_url": "https://huggingface.co/inclusionAI/LLaDA-Image-Turbo",
   "slots": [

examples/server/recipes/llada-image.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "llada-image",
   "name": "LLaDA Image (base)",
+  "size_multiple": 16,
   "description": "Use matching base transformer, text encoder and connectors; do not mix them with Turbo components. Editing needs connectors with SigVQ and 1024x1024 output.",
   "model_url": "https://huggingface.co/inclusionAI/LLaDA-Image",
   "slots": [

examples/server/recipes/ltx23-dev.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "ltx23-dev",
   "name": "LTX 2.3 dev",
+  "size_multiple": 32,
   "description": "Supports text, start-image and start/end-image video. Audio VAE is optional for silent video. Configure latent upscaling separately in generation settings. Optional distillation LoRAs use this recipe; select the LoRA and adjust the sampling settings separately.",
   "model_url": "https://huggingface.co/Lightricks/LTX-2.3",
   "slots": [
@@ -39,6 +40,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 8,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "sampling_method": "euler",
     "video_frames": 33,

examples/server/recipes/ltx23-distilled.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "ltx23-distilled",
   "name": "LTX 2.3 Distilled 1.1",
+  "size_multiple": 32,
   "description": "Supports text, start-image and start/end-image video. Audio VAE is optional for silent video. Configure latent upscaling separately in generation settings.",
   "model_url": "https://huggingface.co/Lightricks/LTX-2.3",
   "slots": [
@@ -39,6 +40,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 8,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "sampling_method": "euler",
     "video_frames": 121,

examples/server/recipes/ltx25-dev.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "ltx25-dev",
   "name": "LTX 2.5 dev",
+  "size_multiple": 32,
   "description": "Use the conv video VAE and LTX-specific Gemma 4 with bundled projection. Audio VAE is optional for silent video.",
   "model_url": "https://huggingface.co/Lightricks/LTX-2.5",
   "slots": [
@@ -35,6 +36,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 8,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "sampling_method": "euler",
     "fps": 24,

examples/server/recipes/mage-flow-edit.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "mage-flow-edit",
   "name": "Mage Flow Edit",
+  "size_multiple": 16,
   "description": "Supply reference images. Select a separate vision projector when it is not bundled in the text encoder.",
   "model_url": "https://huggingface.co/microsoft/Mage-Flow-Edit",
   "slots": [

examples/server/recipes/mage-flow-turbo.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "mage-flow-turbo",
   "name": "Mage Flow Turbo",
+  "size_multiple": 16,
   "model_url": "https://huggingface.co/microsoft/Mage-Flow-Turbo",
   "slots": [
     {

examples/server/recipes/mage-flow.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "mage-flow",
   "name": "Mage Flow",
+  "size_multiple": 16,
   "model_url": "https://huggingface.co/microsoft/Mage-Flow",
   "slots": [
     {

examples/server/recipes/ming-image.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "ming-image",
   "name": "Ming Image 0.1",
+  "size_multiple": 16,
   "model_url": "https://huggingface.co/inclusionAI/Ming-Image-0.1-Design",
   "slots": [
     {
@@ -36,8 +37,8 @@
   "defaults": {
     "cfg_scale": 1,
     "steps": 12,
-    "width": 2048,
-    "height": 2048,
+    "width": 1024,
+    "height": 1024,
     "sampling_method": "euler"
   },
   "features": [

examples/server/recipes/minimax-h3-fl2va.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "minimax-h3-fl2va",
   "name": "MiniMax H3 FL2VA",
+  "size_multiple": 32,
   "description": "Generate video and audio, optionally with start/end images. Supports full and pruned FL2VA weights. Audio VAE is optional for silent video.",
   "model_url": "https://huggingface.co/MiniMaxAI/MiniMax-H3",
   "slots": [
@@ -35,6 +36,13 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 17,
+    "frame_offset": 5,
+    "frame_rounding": "ceil",
+    "min_frames": 5,
+    "fixed_fps": 24
+  },
   "defaults": {
     "video_frames": 56,
     "fps": 24,

examples/server/recipes/minimax-h3-ref2va.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "minimax-h3-ref2va",
   "name": "MiniMax H3 Ref2VA",
+  "size_multiple": 32,
   "description": "Supply image, video or audio references with compatible Ref2VA weights. Audio VAE is optional for silent video.",
   "model_url": "https://huggingface.co/MiniMaxAI/MiniMax-H3",
   "slots": [
@@ -35,6 +36,13 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 17,
+    "frame_offset": 5,
+    "frame_rounding": "ceil",
+    "min_frames": 5,
+    "fixed_fps": 24
+  },
   "defaults": {
     "video_frames": 56,
     "fps": 24,
@@ -45,7 +53,10 @@
   },
   "features": [
     "flow_shift",
-    "ref_images"
+    "ref_images",
+    "ref_videos",
+    "ref_audios",
+    "ref_video_audios"
   ],
   "disabled_features": [
     "img2img",

examples/server/recipes/minit2i.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "minit2i",
   "name": "MiniT2I",
+  "size_multiple": 1,
   "description": "Pixel-space model; no VAE is required.",
   "model_url": "https://huggingface.co/MiniT2I/MiniT2I",
   "slots": [

examples/server/recipes/moss-image-base.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "moss-image-base",
   "name": "MOSS Image Base",
+  "size_multiple": 16,
   "description": "Select the MOSS-Image-Base transformer with its matching text encoder and FLUX.2 VAE or TAE.",
   "model_url": "https://huggingface.co/OpenMOSS-Team/MOSS-Image-Base",
   "slots": [

examples/server/recipes/moss-image-sft.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "moss-image-sft",
   "name": "MOSS Image SFT",
+  "size_multiple": 16,
   "description": "Select the MOSS-Image-SFT transformer with its matching text encoder and FLUX.2 VAE or TAE.",
   "model_url": "https://huggingface.co/OpenMOSS-Team/MOSS-Image-SFT",
   "slots": [

examples/server/recipes/nitrosd-realism.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "nitrosd-realism",
   "name": "NitroSD Realism",
+  "size_multiple": 64,
   "model_url": "https://huggingface.co/ChenDY/NitroFusion",
   "slots": [
     {

examples/server/recipes/pid-15-flux2.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "pid-15-flux2",
   "name": "PiD 1.5 (Flux.2 VAE)",
+  "size_multiple": 1,
   "description": "Supply a reference image; the first image is used as the source latent. Select a matching Flux.2 PiD checkpoint and VAE.",
   "model_url": "https://huggingface.co/nvidia/PiD",
   "slots": [

examples/server/recipes/pid-15-qwen-image.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "pid-15-qwen-image",
   "name": "PiD 1.5 (Qwen Image VAE)",
+  "size_multiple": 1,
   "description": "Supply a reference image; the first image is used as the source latent. Select a matching Qwen Image PiD checkpoint and VAE. Use the Qwen Image 2D VAE.",
   "model_url": "https://huggingface.co/nvidia/PiD",
   "slots": [

examples/server/recipes/pid-15-sd3.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "pid-15-sd3",
   "name": "PiD 1.5 (SD3 VAE)",
+  "size_multiple": 1,
   "description": "Supply a reference image; the first image is used as the source latent. Select a matching SD3 PiD checkpoint and VAE.",
   "model_url": "https://huggingface.co/nvidia/PiD",
   "slots": [

examples/server/recipes/pid-15.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "pid-15",
   "name": "PiD 1.5 (Flux / Z-Image VAE)",
+  "size_multiple": 1,
   "description": "Supply a reference image; the first image is used as the source latent. Select a Flux/Z-Image PiD checkpoint and matching Flux VAE.",
   "model_url": "https://huggingface.co/nvidia/PiD",
   "slots": [

examples/server/recipes/pid-flux2.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "pid-flux2",
   "name": "PiD (Flux.2 VAE)",
+  "size_multiple": 1,
   "description": "Supply a reference image; the first image is used as the source latent. Select a matching Flux.2 PiD checkpoint and VAE.",
   "model_url": "https://huggingface.co/nvidia/PiD",
   "slots": [

examples/server/recipes/pid-qwen-image.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "pid-qwen-image",
   "name": "PiD (Qwen Image VAE)",
+  "size_multiple": 1,
   "description": "Supply a reference image; the first image is used as the source latent. Select a matching Qwen Image PiD checkpoint and VAE. Use the Qwen Image 2D VAE.",
   "model_url": "https://huggingface.co/nvidia/PiD",
   "slots": [

examples/server/recipes/pid-sd3.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "pid-sd3",
   "name": "PiD (SD3 VAE)",
+  "size_multiple": 1,
   "description": "Supply a reference image; the first image is used as the source latent. Select a matching SD3 PiD checkpoint and VAE.",
   "model_url": "https://huggingface.co/nvidia/PiD",
   "slots": [

examples/server/recipes/pid.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "pid",
   "name": "PiD (Flux / Z-Image VAE)",
+  "size_multiple": 1,
   "description": "Supply a reference image; the first image is used as the source latent. Select a Flux/Z-Image PiD checkpoint and matching Flux VAE.",
   "model_url": "https://huggingface.co/nvidia/PiD",
   "slots": [

examples/server/recipes/pixart-alpha-1024.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "pixart-alpha-1024",
   "name": "PixArt Alpha XL 1024",
+  "size_multiple": 16,
   "description": "Use the matching PixArt Alpha transformer and SD 1.x-style VAE. Resolution/aspect-ratio micro-conditioning is not currently applied by the runtime.",
   "model_url": "https://huggingface.co/PixArt-alpha/PixArt-XL-2-1024-MS",
   "slots": [

examples/server/recipes/pixart-alpha-512.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "pixart-alpha-512",
   "name": "PixArt Alpha XL 512",
+  "size_multiple": 16,
   "description": "Use the matching PixArt Alpha transformer and SD 1.x-style VAE. Resolution/aspect-ratio micro-conditioning is not currently applied by the runtime.",
   "model_url": "https://huggingface.co/PixArt-alpha/PixArt-XL-2-512x512",
   "slots": [

examples/server/recipes/pixart-sigma.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "pixart-sigma",
   "name": "PixArt Sigma XL 1024",
+  "size_multiple": 16,
   "model_url": "https://huggingface.co/PixArt-alpha/PixArt-Sigma-XL-2-1024-MS",
   "slots": [
     {

examples/server/recipes/qwen-image-21-turbo.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "qwen-image-21-turbo",
   "name": "Qwen Image 2.1 Turbo",
+  "size_multiple": 32,
   "description": "Use the standalone Turbo diffusion weights for generation and editing with the fixed 8-step schedule. Select the Qwen Image 2.1 VAE or TAEQI2.1, and supply a vision projector when needed.",
   "model_url": "https://huggingface.co/Qwen/Qwen-Image-2.1-Turbo",
   "slots": [
@@ -40,8 +41,8 @@
     "sampling_method": "euler",
     "cfg_scale": 1,
     "steps": 8,
-    "width": 2048,
-    "height": 2048,
+    "width": 1024,
+    "height": 1024,
     "custom_sigmas": [
       1.0,
       0.978453,

examples/server/recipes/qwen-image-21.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "qwen-image-21",
   "name": "Qwen Image 2.1",
+  "size_multiple": 32,
   "description": "Supports generation and editing. Select the full Qwen Image 2.1 VAE or TAEQI2.1. Supply a separate vision projector when needed. Select acceleration LoRAs and adjust their settings separately.",
   "model_url": "https://huggingface.co/Qwen/Qwen-Image-2.1",
   "slots": [
@@ -40,8 +41,8 @@
     "sampling_method": "euler",
     "cfg_scale": 1,
     "steps": 20,
-    "width": 2048,
-    "height": 2048
+    "width": 1024,
+    "height": 1024
   },
   "features": [
     "flow_shift",

examples/server/recipes/qwen-image.json

@@ -35,8 +35,8 @@
     "flow_shift": 3.0,
     "cfg_scale": 4,
     "steps": 20,
-    "width": 1328,
-    "height": 1328
+    "width": 1024,
+    "height": 1024
   },
   "features": [
     "flow_shift"

examples/server/recipes/sd-turbo.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sd-turbo",
   "name": "SD Turbo",
+  "size_multiple": 64,
   "description": "Select a single-file SD-Turbo checkpoint. Uses one step with CFG disabled (scale 1 in sd.cpp); use more steps for partial image-to-image denoising.",
   "model_url": "https://huggingface.co/stabilityai/sd-turbo",
   "slots": [

examples/server/recipes/sd1-checkpoint.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sd1-checkpoint",
   "name": "SD 1.x",
+  "size_multiple": 64,
   "description": "Supports SD 1.x, compatible LCM checkpoints and layer-count variants. For LCM weights, select LCM sampling and adjust steps/CFG. Tiny SD Diffusers directories must first be converted into a single-file checkpoint.",
   "model_url": "https://huggingface.co/CompVis/stable-diffusion-v1-4",
   "slots": [

examples/server/recipes/sd1-controlnet.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sd1-controlnet",
   "name": "SD 1.5 ControlNet",
+  "size_multiple": 64,
   "description": "Supply a control image; input image and mask are optional for image-to-image or inpainting.",
   "model_url": "https://huggingface.co/lllyasviel/control_v11p_sd15_canny",
   "slots": [

examples/server/recipes/sd1-inpainting.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sd1-inpainting",
   "name": "SD 1.5 Inpainting",
+  "size_multiple": 64,
   "description": "Select an inpainting checkpoint, then supply an input image and mask.",
   "slots": [
     {

examples/server/recipes/sd1-ip-adapter.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sd1-ip-adapter",
   "name": "SD 1.5 IP-Adapter / Plus",
+  "size_multiple": 64,
   "description": "Select matching SD 1.5 IP-Adapter or Plus weights and CLIP Vision H, then supply an IP-Adapter image.",
   "model_url": "https://huggingface.co/h94/IP-Adapter",
   "slots": [

examples/server/recipes/sd2-512-checkpoint.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sd2-512-checkpoint",
   "name": "SD 2.1 512",
+  "size_multiple": 64,
   "description": "Supports SD 2.x 512 and compatible layer-count variants. Tiny SD2 Diffusers directories must first be converted into a single-file checkpoint.",
   "slots": [
     {

examples/server/recipes/sd2-768-checkpoint.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sd2-768-checkpoint",
   "name": "SD 2.1 768",
+  "size_multiple": 64,
   "slots": [
     {
       "id": "model",

examples/server/recipes/sd2-inpainting.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sd2-inpainting",
   "name": "SD 2 Inpainting",
+  "size_multiple": 64,
   "description": "Select an inpainting checkpoint, then supply Init Image and Mask Image.",
   "slots": [
     {

examples/server/recipes/sdxl-checkpoint.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sdxl-checkpoint",
   "name": "SDXL",
+  "size_multiple": 64,
   "description": "Supports SDXL and compatible reduced-depth checkpoints such as SSD-1B and Segmind Vega.",
   "model_url": "https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0",
   "slots": [

examples/server/recipes/sdxl-controlnet.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sdxl-controlnet",
   "name": "SDXL ControlNet",
+  "size_multiple": 64,
   "description": "Select a compatible SDXL ControlNet and supply a preprocessed control image. Input image and mask are optional for image-to-image or inpainting.",
   "model_url": "https://huggingface.co/diffusers/controlnet-canny-sdxl-1.0",
   "slots": [

examples/server/recipes/sdxl-inpainting.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sdxl-inpainting",
   "name": "SDXL Inpainting",
+  "size_multiple": 64,
   "description": "Select an inpainting checkpoint, then supply Init Image and Mask Image.",
   "model_url": "https://huggingface.co/diffusers/stable-diffusion-xl-1.0-inpainting-0.1",
   "slots": [

examples/server/recipes/sdxl-ip-adapter.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sdxl-ip-adapter",
   "name": "SDXL IP-Adapter / Plus",
+  "size_multiple": 64,
   "description": "Select matching SDXL ViT-H IP-Adapter or Plus weights and CLIP Vision H, then supply an IP-Adapter image.",
   "model_url": "https://huggingface.co/h94/IP-Adapter",
   "slots": [

examples/server/recipes/sdxl-photomaker.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sdxl-photomaker",
   "name": "SDXL PhotoMaker",
+  "size_multiple": 64,
   "description": "Select PhotoMaker v1 and supply identity images. Include the img trigger in the prompt.",
   "model_url": "https://huggingface.co/TencentARC/PhotoMaker",
   "slots": [

examples/server/recipes/sdxl-turbo.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sdxl-turbo",
   "name": "SDXL Turbo",
+  "size_multiple": 64,
   "model_url": "https://huggingface.co/stabilityai/sdxl-turbo",
   "slots": [
     {

examples/server/recipes/sdxs.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sdxs",
   "name": "SDXS",
+  "size_multiple": 64,
   "model_url": "https://huggingface.co/IDKiro/sdxs-512-dreamshaper",
   "slots": [
     {

examples/server/recipes/sefi-image.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sefi-image",
   "name": "SeFi Image 5B RL",
+  "size_multiple": 16,
   "model_url": "https://huggingface.co/SeFi-Image/SeFi-Image-5B-Base",
   "slots": [
     {

examples/server/recipes/sensenova-u15.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "sensenova-u15",
   "name": "SenseNova U1.5",
+  "size_multiple": 32,
   "description": "Bundled MoT checkpoint.",
   "model_url": "https://huggingface.co/sensenova/SenseNova-U1.5-8B-MoT",
   "slots": [
@@ -14,8 +15,8 @@
     "flow_shift": 3.0,
     "cfg_scale": 4,
     "steps": 20,
-    "width": 2048,
-    "height": 2048,
+    "width": 1024,
+    "height": 1024,
     "sampling_method": "euler"
   },
   "features": [

examples/server/recipes/wan21-flf2v.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "wan21-flf2v",
   "name": "Wan 2.1 FLF2V 14B",
+  "size_multiple": 16,
   "description": "Select FLF2V weights and supply both start and end images.",
   "model_url": "https://huggingface.co/Wan-AI/Wan2.1-FLF2V-14B-720P",
   "slots": [
@@ -35,6 +36,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 4,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "sampling_method": "euler",
     "flow_shift": 16,

examples/server/recipes/wan21-i2v.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "wan21-i2v",
   "name": "Wan 2.1 I2V",
+  "size_multiple": 16,
   "description": "Image-to-video with Wan 2.1, compatible SkyReels V2 weights, or a single Wan 2.2 I2V expert. CLIP Vision is required by Wan 2.1 I2V and is unused by Wan 2.2. Defaults target Wan 2.1 480p.",
   "model_url": "https://huggingface.co/Wan-AI/Wan2.1-I2V-14B-480P",
   "slots": [
@@ -35,6 +36,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 4,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "sampling_method": "euler",
     "video_frames": 33,

examples/server/recipes/wan21-t2v.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "wan21-t2v",
   "name": "Wan 2.1 T2V",
+  "size_multiple": 16,
   "description": "Text-to-video with Wan 2.1 or a compatible single Wan 2.2 T2V expert. Defaults target Wan 2.1 14B; adjust settings for the selected weights and acceleration LoRAs.",
   "model_url": "https://huggingface.co/Wan-AI/Wan2.1-T2V-14B",
   "slots": [
@@ -30,6 +31,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 4,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "sampling_method": "euler",
     "video_frames": 33,

examples/server/recipes/wan21-vace.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "wan21-vace",
   "name": "Wan 2.1 VACE 1.3B / 14B",
+  "size_multiple": 16,
   "description": "Select matching VACE weights. Supports text, reference images and control video through Video Generation.",
   "model_url": "https://huggingface.co/Wan-AI/Wan2.1-VACE-14B",
   "slots": [
@@ -30,6 +31,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 4,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "sampling_method": "euler",
     "video_frames": 33,

examples/server/recipes/wan22-i2v.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "wan22-i2v",
   "name": "Wan 2.2 I2V A14B",
+  "size_multiple": 16,
   "description": "Image-to-video. Select a start image in Video Generation after loading. LightX2V acceleration LoRAs must match the I2V low- and high-noise models; configure LoRAs and sampling separately.",
   "model_url": "https://huggingface.co/Wan-AI/Wan2.2-I2V-A14B",
   "slots": [
@@ -35,6 +36,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 4,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "sampling_method": "euler",
     "video_frames": 33,

examples/server/recipes/wan22-s2v.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "wan22-s2v",
   "name": "Wan 2.2 S2V 14B",
+  "size_multiple": 16,
   "description": "Supply a start image and reference audio through Video Generation.",
   "model_url": "https://huggingface.co/Wan-AI/Wan2.2-S2V-14B",
   "slots": [
@@ -35,6 +36,13 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 4,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1,
+    "fixed_fps": 16
+  },
   "defaults": {
     "sampling_method": "euler",
     "flow_shift": 3,
@@ -44,7 +52,8 @@
   },
   "features": [
     "flow_shift",
-    "init_image"
+    "init_image",
+    "ref_audios"
   ],
   "disabled_features": [
     "img2img",

examples/server/recipes/wan22-t2v.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "wan22-t2v",
   "name": "Wan 2.2 T2V A14B",
+  "size_multiple": 16,
   "description": "Text-to-video. Use Video Generation after loading. LightX2V acceleration LoRAs must match the T2V low- and high-noise models; configure LoRAs and sampling separately.",
   "model_url": "https://huggingface.co/Wan-AI/Wan2.2-T2V-A14B",
   "slots": [
@@ -35,6 +36,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 4,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "sampling_method": "euler",
     "video_frames": 33,

examples/server/recipes/wan22-ti2v-5b.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "wan22-ti2v-5b",
   "name": "Wan 2.2 TI2V 5B",
+  "size_multiple": 32,
   "description": "Use Wan 2.2 VAE. Supports text, start-image and start/end-image conditioning.",
   "model_url": "https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B",
   "slots": [
@@ -30,6 +31,12 @@
       "tae"
     ]
   ],
+  "video_constraints": {
+    "frame_step": 4,
+    "frame_offset": 1,
+    "frame_rounding": "floor",
+    "min_frames": 1
+  },
   "defaults": {
     "sampling_method": "euler",
     "video_frames": 33,

examples/server/recipes/z-image-l2p.json

@@ -2,6 +2,7 @@
   "schema_version": 1,
   "id": "z-image-l2p",
   "name": "Z-Image L2P",
+  "size_multiple": 16,
   "description": "Pixel-space L2P checkpoint with Qwen 3 4B; no VAE is needed. Use dimensions divisible by 16.",
   "model_url": "https://huggingface.co/zhen-nan/L2P",
   "slots": [

examples/server/routes_sdcpp.cpp

@@ -212,11 +212,15 @@ static json make_img_gen_features_json() {
     };
 }
 
+/** Builds the default feature set advertised for vid_gen requests. */
 static json make_vid_gen_features_json() {
     return {
         {"init_image", true},
         {"end_image", true},
         {"control_frames", true},
+        {"ref_videos", true},
+        {"ref_audios", true},
+        {"ref_video_audios", true},
         {"high_noise_sample_params", true},
         {"lora", true},
         {"vae_tiling", true},
@@ -235,6 +239,7 @@ static json make_preview_modes_json() {
     return modes;
 }
 
+/** Builds the /capabilities payload: model, defaults, limits and feature maps. */
 static json make_capabilities_json(ServerRuntime& runtime) {
     refresh_lora_cache(runtime);
     refresh_upscaler_cache(runtime);
@@ -351,8 +356,10 @@ static json make_capabilities_json(ServerRuntime& runtime) {
     const json generation_features = model_generation_features(runtime);
     auto recipe_features           = [&](json features, bool video) {
         features["init_image"] = generation_features.at(video ? "init_image" : "img2img");
-        for (const char* key : {"mask_image", "ref_images", "end_image", "control_frames", "lora", "vae_tiling", "cache", "preview"}) {
-            if (features.contains(key) || std::string(key) == "ref_images")
+        for (const char* key : {"mask_image", "ref_images", "end_image", "control_frames", "ref_videos", "ref_audios", "ref_video_audios", "lora", "vae_tiling", "cache", "preview"}) {
+            if (features.contains(key) || std::string(key) == "ref_images" ||
+                (video && (std::string(key) == "ref_videos" || std::string(key) == "ref_audios" ||
+                           std::string(key) == "ref_video_audios")))
                 features[key] = generation_features.at(key);
         }
         if (video)

src/model/diffusion/moss_image.hpp

@@ -185,7 +185,7 @@ namespace MOSSImage {
                     ggml_build_forward_expand(ctx->graph, parts[i]);
                 }
             }
-            x         = ggml_concat(c, x, text, 1);
+            x = ggml_concat(c, x, text, 1);
             for (int64_t i = 0; i < config.depth; ++i) {
                 if (std::find(skip.begin(), skip.end(), i) == skip.end()) {
                     x = std::dynamic_pointer_cast<Block>(blocks["layers." + std::to_string(i)])->forward(ctx, x, pe, parts);

src/name_conversion.cpp

@@ -708,6 +708,24 @@ std::string convert_diffusers_dit_to_original_flux(std::string name) {
     return name;
 }
 
+static std::string convert_diffusers_dit_to_original_flux3(std::string name) {
+    static const std::vector<std::pair<std::string, std::string>> prefix_map = {
+        {"single_transformer_blocks.", "single_blocks."},
+        {"mode_blocks.", "content_mode_blocks."},
+    };
+    static const std::vector<std::pair<std::string, std::string>> name_map = {
+        {".attn.to_q.", ".q_proj."},
+        {".attn.to_k.", ".k_proj."},
+        {".attn.to_v.", ".v_proj."},
+        {".attn.to_out.0.", ".attn_out."},
+        {".ff.linear_in.", ".mlp_in."},
+        {".ff.linear_out.", ".mlp_out."},
+    };
+    replace_with_prefix_map(name, prefix_map);
+    replace_with_name_map(name, name_map);
+    return name;
+}
+
 std::string convert_hunyuan_video_to_original_flux(std::string name) {
     int num_layers        = 54;
     int num_single_layers = 0;
@@ -1077,6 +1095,8 @@ std::string convert_diffusion_model_name(std::string name, std::string prefix, S
         name = convert_diffusers_dit_to_original_sd3(name);
     } else if (sd_version_is_flux(version) || sd_version_is_flux2(version) || sd_version_is_longcat(version) || sd_version_is_sefi_image(version)) {
         name = convert_diffusers_dit_to_original_flux(name);
+    } else if (sd_version_is_flux3(version)) {
+        name = convert_diffusers_dit_to_original_flux3(name);
     } else if (sd_version_is_hunyuan_video(version)) {
         name = convert_hunyuan_video_to_original_flux(name);
     } else if (version == VERSION_MOSS_IMAGE) {

src/pipeline/image.cpp

@@ -314,8 +314,8 @@ namespace sd::pipeline {
         }
 
         bool skip_vae_encoding = (sd->version == VERSION_HIDREAM_O1) || !(
-            ref_image_params.pass_to_dit || ref_image_params.vlm_use_vae_input ||
-            sd_version_is_unet_edit(sd->version) || sd_version_is_pid(sd->version));
+                                                                            ref_image_params.pass_to_dit || ref_image_params.vlm_use_vae_input ||
+                                                                            sd_version_is_unet_edit(sd->version) || sd_version_is_pid(sd->version));
 
         std::vector<sd::Tensor<float>> ref_latents;
         for (size_t i = 0; i < ref_images.size(); i++) {

src/runtime/image_preprocess.cpp

@@ -379,8 +379,13 @@ namespace sd {
         if (target == ImageTarget::Mask && has_init_transform_) {
             auto options = resolve_options(target, index);
             if (image.width != init_transform_.source_width || image.height != init_transform_.source_height) {
-                fail("mask and init source dimensions must match");
-                return {};
+                ImagePreprocessor alignment;
+                tensor = alignment.apply_geometry(tensor, target, index, init_transform_.source_width, init_transform_.source_height,
+                                                  ImageResizeMode::CropResize, ops::InterpolateMode::NearestExact);
+                if (tensor.empty()) {
+                    fail("could not align mask to init source dimensions");
+                    return {};
+                }
             }
             bool geometry_override = options.count("width") || options.count("anchor") ||
                                      (options.count("mode") && options.at("mode") != "auto");