feat(animatediff): support img2video via --init-img (#1789)

docs/animatediff.md

@@ -134,6 +134,25 @@ sd-cli -M vid_gen --model realisticVisionV60B1.safetensors \
        -p "close up photo of a rabbit ...<lora:v3_sd15_adapter:1.0>" ...
 ```
 
+## img2video
+
+Pass a pre-rendered image via `-i / --init-img` to animate FROM it. All N output frames start from the encoded init latent, then per-frame noise is added at `--strength`. Character identity, composition, and quality are anchored by the init image; the motion module adds subtle motion on top.
+
+Left: init image rendered with `-M img_gen`. Right: 8-frame vid_gen output.
+
+<img src="../assets/animatediff/img2video_demo.gif" width="512"/>
+
+```
+sd-cli -M img_gen ... -o init.png                  # any high-quality still
+sd-cli -M vid_gen --motion-module mm_sd15_v3.safetensors \
+    -i init.png --strength 0.75 \
+    --cfg-scale 7.0 --sampling-method euler --scheduler karras \
+    -H 512 -W 512 --video-frames 8 --steps 25 -s 42 \
+    -p "..." -o out.avi
+```
+
+`--strength` controls how far the motion module is allowed to deviate from the init image (higher = more motion, lower = more static).
+
 ## Notes
 
 - The motion module was trained at `video_length=16`. Running with

src/stable-diffusion.cpp

@@ -4687,8 +4687,16 @@ static std::optional<ImageGenerationLatents> prepare_image_generation_latents(sd
         init_latent.dim() >= 4 && init_latent.shape()[3] == 1) {
         int n_frames = sd_ctx->sd->animatediff_num_frames;
         std::vector<int64_t> shape(init_latent.shape().begin(), init_latent.shape().end());
-        shape[3]    = n_frames;
+        shape[3] = n_frames;
-        init_latent = sd::Tensor<float>(std::move(shape));  // zero-filled batch of N frames; per-frame noise is generated later via randn_like.
+        if (!init_image_tensor.empty()) {
+            sd::Tensor<float> replicated(shape);
+            for (int f = 0; f < n_frames; ++f) {
+                sd::ops::slice_assign(&replicated, 3, f, f + 1, init_latent);
+            }
+            init_latent = std::move(replicated);
+        } else {
+            init_latent = sd::Tensor<float>(std::move(shape));
+        }
     }
 
     if (!control_image_tensor.empty()) {
@@ -6132,6 +6140,7 @@ static bool generate_animatediff_video(sd_ctx_t* sd_ctx,
     img_gen_params.height            = sd_vid_gen_params->height;
     img_gen_params.sample_params     = sd_vid_gen_params->sample_params;
     img_gen_params.strength          = sd_vid_gen_params->strength;
+    img_gen_params.init_image        = sd_vid_gen_params->init_image;
     img_gen_params.seed              = sd_vid_gen_params->seed;
     img_gen_params.batch_count       = 1;
     img_gen_params.control_strength  = 1.0f;