Video to depth
Turn your video into a grayscale depth MP4. Keep the original resolution, every frame, and the full duration. No settings needed.
Turn choreography, body movement, and camera motion into a grayscale reference for your next video.
Original resolution, frame timing, and full duration. Choose a video and convert — no settings needed.
- Original resolution
- Every original frame
- Full video duration
First use downloads a model of about 50 MB. Later runs can reuse the browser cache. Processing speed depends on your device.
A motion reference. Ready for a new scene.
- 1Choose a video. Its original resolution, frames, and duration are kept.
- 2Extract depth locally, then preview the original and the depth video.
- 3Download your MP4 and use it as a motion reference in a compatible video workflow.
Before you start
Do my files leave my device?
No. These tools process and download files inside this browser. The video engine is downloaded on first use, but your media is not uploaded. Closing or refreshing the page clears the current files and results.
What does a depth video preserve?
Brightness represents estimated relative distance, not the original colors. Body movement, silhouettes, scene geometry, and camera motion can remain. This is a motion reference, not a measured 3D reconstruction.
Can I use this for an AI dance video?
Yes. Extract the reference movement, create an image of your new character and scene, then use the depth clip and character image in a video model that accepts video references. Check that model's format and duration limits first; generation is a separate step.
Which browsers support depth extraction?
Use desktop Chrome or Edge with WebGPU and hardware acceleration enabled. First use downloads a model of about 50 MB. Your video stays on your device. Low-memory devices have a 40 MiB input limit.
Does the output match my video?
Yes. The entire video is converted at its original resolution, with every frame and its original timing, including variable frame rates. Output is a silent H.264 MP4. Depth detail is estimated by the model; matching dimensions does not make it a measured 3D reconstruction.
Which videos can I convert?
Choose an MP4, MOV, or WebM up to five minutes, 1080p, and 100 MiB (40 MiB on low-memory devices). Longer videos take more time. If the original dimensions, timing, or memory needs are unsupported, the tool reports an error instead of reducing quality or cutting the video.
Why do edges or brightness sometimes flicker?
Depth is estimated one frame at a time. Fast movement, occlusion, reflective surfaces, and scene cuts can cause flicker or unstable edges. Clear, continuous shots usually work best.
Does depth conversion anonymize the video?
No. Depth removes original color and texture, but body shape, movement, and scene structure may still identify someone. It does not guarantee anonymity or a downstream model's acceptance.