Video object removal is one of the most powerful yet frustrating AI workflows today.
Most existing solutions are:
- Cloud-based → slow uploads, privacy concerns, video length limits.
- Research demos → impressive results, but painful to install or scale.
- Image-only tools → fine for photos, but fail on videos (flicker, no temporal consistency).
VideoVanish bridges this gap by offering:
- A local-first workflow — no uploads, no hidden costs.
- A simple installer + GUI — no command line needed.
- State-of-the-art AI models (DiffuEraser, SAM2) built-in.
- Support for long videos with chunking, overlap, and blending.
- Both GUI for ease of use and CLI for automation.
👉 In short: professional-grade AI video inpainting, without the research-paper headaches.
- Install Miniconda with latest installer (choose all defaults).
- Download VideoVanish (main.zip) and extract it anywhere.
- Double-click
windows_install.bat. - Double-click
start_videovanish.batto launch.
git clone https://github.com/calledit/VideoVanish
cd VideoVanish
./install_videovanish.sh
conda activate videovanish
python videovanish.pyMac version is work in progress and videos may not play properly on Mac.
VideoVanish is intended to be a user-friendly tool for state-of-the-art video object removal and masking.
- Simple installer, no command line required.
- Basic GUI for video editing (timeline + preview).
- Not browser-based, and does not use ComfyUI.
- Load color video + optional mask video.
- If no mask video, create one in GUI using SAM2:
- Left click = add point
- Right click = remove point/box
- Drag = define area
- Click Generate Mask to build mask video
- With both color + mask, press Vanish → uses DiffuEraser to remove objects.
- Adjustable inference resolution.
- Result is rescaled and blended into original video.
- Dependencies (SAM2 + DiffuEraser models) installed automatically.
- Command-line support for automation.
- Hide console window on startup (many users find it annoying).
- Keep a way to view weight download progress (inline console window or GUI download manager).
- Split videos into chunks with overlapping frames to reduce vram requirements. Overlaps should be blended during stitching.
⚠️ Perfect blending may not always be possible.