Merge branch 'experimental'

SKILL.md

@@ -128,6 +128,9 @@ python3 videogen --auto --prompt "..." --allow-bigger-models
 # Specify offload strategy explicitly
 python3 videogen --auto --prompt "..." --offload_strategy sequential --vram_limit 16
 
+# NEW: Balanced strategy - maximizes VRAM usage, only offloads if necessary
+python3 videogen --auto --prompt "..." --offload_strategy balanced
+
 # Limit VRAM usage
 python3 videogen --model wan_14b_t2v --prompt "..." --vram_limit 16
 
@@ -492,7 +495,8 @@ VideoGen creates these output files:
 
 2. **Out of memory**
    - Use a smaller model
-   - Add `--offload_strategy sequential`
+   - Add `--offload_strategy sequential` (for very large models)
+   - Use `--offload_strategy balanced` (recommended - maximizes VRAM usage)
    - Reduce resolution with `--width` and `--height`
 
 3. **Pipeline class not found**