first stab at adding video

Dockerfile

@@ -4,6 +4,9 @@ ENV DEBIAN_FRONTEND=noninteractive
 ENV PYTHONUNBUFFERED=1
 # HuggingFace model cache — mounted as a volume so models persist across runs
 ENV HF_HOME=/cache/huggingface
+# LoRA directory — users drop .safetensors files here and reference them
+# from config.yml::video.loras. Bind-mounted via docker-compose.
+ENV LORA_DIR=/cache/loras
 
 RUN apt-get update && apt-get install -y \
     python3.11 \
@@ -38,6 +41,26 @@ RUN python3.11 -m pip install --no-cache-dir -r requirements.txt
 # Pre-download the spacy model that kokoro needs at runtime
 RUN python3.11 -m spacy download en_core_web_sm
 
+# --- Optional: avatar video stack -------------------------------------------
+# These are heavy installs; keep them after the core deps so rebuilds only
+# redo this layer when ONLY the video stack changes. If you don't plan to
+# use config.video.enabled=true, you can comment this block out to speed
+# up builds and shrink the image.
+#
+# LightX2V (Wan2.2-Lightning inference framework) — installed from source
+# since there is no stable PyPI release yet.
+RUN python3.11 -m pip install --no-cache-dir \
+    "git+https://github.com/ModelTC/LightX2V.git" || \
+    echo "LightX2V install failed — config.video.enabled must stay false until fixed"
+#
+# MuseTalk (audio-driven lip-sync) — same story.
+RUN python3.11 -m pip install --no-cache-dir \
+    "git+https://github.com/TMElyralab/MuseTalk.git" || \
+    echo "MuseTalk install failed — config.video.enabled must stay false until fixed"
+#
+# LoRA directory (user drops .safetensors here; bind-mounted in compose).
+RUN mkdir -p /cache/loras
+
 COPY . .
 
 EXPOSE 8000