- In `image.py`, added a final batch bypass (`advance_check`) to disable face detection and cropping when the current batch completes the remaining video sequence. This prevents the generation of single-frame dummy batches at the end.
- In `video.py` and `loop.py`, updated `AutoLoopCalculator`, `AutoLoopCalculatorWan`, and `AutoLoopCalculatorLTX` to use an `absorption_threshold = equitable_target * 1.5`. This forces the calculator to absorb up to 50% extra remaining frames in a single batch, eliminating micro-cycles.
This commit patches the `SessionImageSender`, `AutoLoopCalculatorWan`,
and `AutoLoopCalculatorLTX` nodes to prevent infinite loops caused by
extreme truncation at the end of video sequences.
- `image.py`: Ensures `SessionImageSender` accepts the entire batch if no
faces are detected, rather than defaulting to frame 0, which previously
trapped the workflow in a 1-frame micro-cycle loop.
- `video.py`: Updates `AutoLoopCalculatorWan` to always round the final
chunk size up to the nearest multiple of 4, removing the strict
cropping logic that broke the 3D VAE rule.
- `loop.py`: Updates `AutoLoopCalculatorLTX` to always round the final
chunk size up to the nearest `8n + 1` value, removing the strict
cropping logic that broke the DiT rule.
- Created `MeisoftVRAMDefragmenter` in `vram_node.py` as a passthrough to force model unloading, Python garbage collection, and CUDA cache defragmentation.
- Registered the new node in `__init__.py`.
- Added light memory cleanup on intermediate batch loops in `loop.py`, alongside a heavy cleanup (including ctypes `malloc_trim`) at the end.
- Added `unload_detector_after_analysis` boolean flag to `VideoAnalyzerWithAudio` in `video.py` to allow releasing the BBOX detector model immediately after use.
Added `ConditionalAudioRouter` to `tools.py` with `torch` import to generate an empty audio tensor for bypassing a LipSync node during intermediate loops without breaking it. Registered node to `__init__.py`.
- Updated `SessionImageSender` to conditionally save keyframe only when not in the final cycle.
- Updated `IncrementalVideoStitcher` to return full concatenated tensor when in the final cycle, avoiding dummy frame output and blocking downstream nodes.
- Expose `IS_FINAL_CYCLE` boolean from `IncrementalVideoStitcher` to UI for downstream conditionals.
- Replaces LoadVideoWithSourceAudio with VideoAnalyzerWithAudio (the "Explorer").
- Introduces OpenCV-based frame scanning to detect sharp and frontal faces.
- Updates AutoLoopCalculator (the "Brain") to utilize safe_faces_list for asymmetric intelligent cuts.
- Updates __init__.py to register the new node correctly with the corresponding display name mapping ("🕵️ Video Analyzer + Audio").
- Introduced global state memory (`global_source_frame_count`, `global_target_frames`, `global_stride`) in `loop.py`.
- Removed visual `source_frame_count` wire from `AutoLoopCalculator` in `video.py`
- Upgraded `AutoLoopCalculator` to execute as a blind "Explorer" in Cycle 0.
- Updated `LoadVideoWithSourceAudio` to intercept real `source_frame_count` directly from VHS payloads.
- Added dynamic calculation of total cycles based on intercepted frames at the end of Cycle 0.
- Se actualizaron las descripciones en README.md y README_EN.md para reflejar las nuevas capacidades de LoadVideoWithSourceAudio (widget UI intacto, upload directo) y el Stitcher (INPUT_IS_LIST=False, passthrough directo de audio al último frame).
- Se restauraron los scripts de testing en `tests/` que se habían eliminado accidentalmente.
- Se reimplementó `LoadVideoWithSourceAudio` para envolver adecuadamente `VHS_LoadVideo`, añadiendo la directiva de UI "video_upload", y métodos para la verificación de cambios (IS_CHANGED, VALIDATE_INPUTS) mientras se conserva el widget de UI.
- Se ha expuesto explícitamente `INPUT_IS_LIST = False` en `IncrementalVideoStitcher` para forzar el tratamiento de imágenes como tensores puros.
- El passthrough de audio en `IncrementalVideoStitcher` ya funciona correctamente de forma predeterminada sin cambios adicionales.
- Se agregó el nuevo alias de nodo en `__init__.py` para su registro en ComfyUI.
Changes:
- Modified IncrementalVideoStitcher to process IMAGE tensors directly instead of MP4 lists.
- Implemented RAM/VRAM saving mechanism by saving tensor state dynamically to `.pt` files and cleaning them manually via `gc`.
- Introduced LoadVideoWithSourceAudio mapping to extract complete original tracks dynamically.
- Created `video.py` introducing the `FFmpegVideoStitcher` node to concatenate video chunks generated via sequential batch loops.
- Registered the node into ComfyUI's standard mapping dictionaries in `__init__.py`.
- Added check for FFmpeg installation using `shutil.which` and raising a ComfyUI-friendly error if it's missing.
- Updated node output file saving logic to utilize ComfyUI's native `folder_paths.get_output_directory()`.
- Revised `README.md` and `README_ES.md` with explicit instructions on how to set up the workflow (VHS Combine integration) and provided warnings emphasizing the need for `easy cleanGpuUsed` after VAE decoders to prevent PyTorch OOM exceptions during iterative batching loops.
Meisoftcoltd