Depth Video Studio — Local AI Video Depth Map Generator
Private by design
Your footage never leaves the browser
The complete media pipeline runs locally. There is no upload endpoint, account, or database.
Workflow
From source clip to depth pass
Choose the model and output controls that fit the clip and the available hardware.
- 01
Choose a short clip
Select a decodable MP4 or WebM up to 60 seconds and 200 MB.
- 02
Generate locally
The selected Depth Anything V2 model runs through WebGPU or WASM while real depth frames appear in the preview.
- 03
Review and download
Play both videos together, seek in sync, then download the audio-free MP4 or WebM.
Production use
A clean depth pass for AI comics and motion workflows
Use the grayscale result as spatial guidance for parallax, relighting, compositing, or AI manhua scene preparation. Always review edges and fast motion before production use.
FAQ
Before you process a clip
Limits, network behavior, output quality, and browser support for this local MVP.
Does my video get uploaded?
No. Decode, depth estimation, preview, and encoding happen in your browser. The site has no media upload API.
Why does the first run need internet access?
The selected model is downloaded from Hugging Face. Browser cache normally avoids downloading it again.
What files can I use?
Use a genuinely decodable MP4 or WebM up to 60 seconds, 200 MB, and 1920×1080-equivalent pixels. A renamed extension is not enough.
Why is the output different from the source?
The depth result is a lossy, audio-free MP4 or WebM at your selected frame rate and size. It is an AI estimate, not measured geometry.
Which browsers work?
The MVP targets current desktop Chrome and Edge with WebGPU or WASM, WebCodecs, and H.264, VP9, or VP8 encoding. Mobile conversion is disabled.