Video Depth Anything · online
Video Depth Map Generator
Upload a video and get a depth map video back. Video Depth Anything, a CVPR 2025 model built to keep depth steady from frame to frame, turns every frame into grayscale depth where brighter means closer. Use it in After Effects, DaVinci Resolve, Blender or ControlNet. No GPU or install.
Make a depth map- MP4 or MOV, up to 60 s
- Grayscale + 4 colormaps
- MP4 output up to 1080p
3.4 credits per second of video, charged on the exact length and rounded up once
Choose a video to continue.
See what a video depth map looks like
The same 6-second clip before and after Video Depth Anything. Sample clip from fal's model page.
What is a video depth map?
A video depth map is a version of a video where each pixel's brightness shows how far that point is from the camera instead of its colour. In the grayscale output here, white is nearest and black is farthest. Editors use it as a matte for fog, depth of field and relighting, and AI video tools use it to keep the 3D layout of a scene while restyling it.
Running a still-image depth model on each frame separately makes the estimated depth jump between frames, which shows up as flicker. Video Depth Anything looks at neighbouring frames together, so depth stays steadier across the clip.
Specs at a glance
| Model | Video Depth Anything (CVPR 2025) on fal, in Large, Base or Small |
|---|---|
| Input | MP4 or MOV, up to 60 seconds, 50 MB and 2,400 frames |
| Output | H.264 MP4 with the same length and frame rate as the input, up to 1080p (larger inputs are scaled down) |
| Colormaps | Grayscale (normalized depth, brighter = closer), Turbo, Inferno, Magma, Viridis |
| Side-by-side | Optional original | depth video in one file |
| Depth type | Relative depth, 8 bits per channel (not metric distance, not 16-bit or EXR) |
| Price | 3.4 credits per second of video, charged on the clip's exact length and rounded up once; failed jobs are refunded |
| Requirements | A browser. No GPU, Python or plugin |
How to make a depth map from a video
Three steps, nothing to install.
1. Upload your clip
Drop in an MP4 or MOV up to 60 seconds. The page reads its length and shows the exact credit cost before anything is charged.
2. Pick a colormap and model size
Grayscale is best for compositing and ControlNet; Turbo, Inferno, Magma and Viridis make depth easier to read. Turn on side-by-side to get the original and the depth map in one video.
3. Download the depth map MP4
Video Depth Anything processes every frame. The result keeps your clip's length and frame rate, so it lines up with the original on your timeline.
Which colormap should you choose?
All five show the same depth; only the colours differ. Each still below is the same frame rendered with that palette.
Grayscale
White is near, black is far. Use it as a matte or depth pass in After Effects, Resolve, Nuke or Blender, and as a ControlNet depth input.
Turbo
A rainbow from warm (near) to cool (far). The easiest to check by eye and to show in presentations.
Inferno
Black through red to yellow. High contrast, and still readable for most people with colour-vision deficiency.
Magma
Black through purple to pale peach. Softer than Inferno; suits dark-themed visuals.
Viridis
Purple through green to yellow. Perceptually uniform, so equal steps in depth look like equal steps in colour.
Use your depth map in After Effects, Resolve, Blender and ComfyUI
Download the grayscale MP4, line it up with the original on your timeline and use it as a luminance matte or control input.
After Effects
Place the depth map above your footage and hide it. Use it as a Luma Matte for a fog or colour layer, or choose it as the depth map layer in Camera Lens Blur for depth of field. Invert it if an effect expects white to be far.
DaVinci Resolve, including the free version
Resolve's built-in Depth Map effect needs Resolve Studio. An external depth map works in the free version too: put it on a track above the footage, then in Fusion use it as a mask for a blur, colour correction or haze layer.
Blender
Load the MP4 as a movie texture and drive a Displace modifier on a subdivided plane for a 2.5D parallax shot, or use it as a mask in the compositor. 8-bit depth can show steps on large displacements, so keep the strength moderate.
ControlNet, ComfyUI and Wan
Feed the grayscale depth video into a depth ControlNet or a video model that accepts depth control, such as Wan or VACE workflows in ComfyUI, to restyle footage while keeping its layout and camera motion. Match the frame rate to your workflow.
3D, parallax and side-by-side
Depth plus the original frame is what 2D-to-3D and parallax tools need. Turn on side-by-side output to get both in one MP4 for review or for tools that take a combined input.
Ways to get a depth map from video, compared
Where this tool fits next to the other common options.
| Option | What you need | Best for |
|---|---|---|
| This page (Video Depth Anything) | A browser and credits | Steady depth passes for any editor or AI workflow without setup |
| Depth Anything V2 per frame (ComfyUI, scripts) | A GPU setup or a hosted node | Single images, or clips where some flicker is acceptable |
| DaVinci Resolve Depth Map effect | DaVinci Resolve Studio (paid) | Working entirely inside Resolve |
| After Effects depth plugins | After Effects plus a plugin licence | Artists who stay inside After Effects |
| Video Depth Anything on your own GPU | Python and a CUDA GPU with plenty of memory | Long clips, raw float depth and batch jobs |
Video depth map FAQ
Limits, formats and how to use the result.
What is a video depth map?
It is a video where each pixel shows distance from the camera instead of colour. In grayscale, white is near and black is far. This tool makes one from your clip with Video Depth Anything, frame by frame, at the same length and frame rate as the original.
How much does a video depth map cost? Is there a free option?
The cost is 3.4 credits per second of video, charged on the clip's exact length and rounded up once, so a 10-second clip costs 34 credits. The exact amount is shown before you generate, and failed jobs are refunded automatically. New accounts get 10 free credits.
Why do depth maps flicker, and does this one?
Flicker comes from running an image depth model on each frame on its own: tiny changes between frames make the estimated depth jump. Video Depth Anything uses neighbouring frames together, which keeps depth much steadier. Fast motion, scene cuts and reflections can still wobble, so trim at cuts for the cleanest result.
Which colormap should I use?
Use Grayscale for compositing, Blender and ControlNet, because software reads brightness as depth. Use Turbo, Inferno, Magma or Viridis when people need to read the depth by eye. All five carry the same depth.
How long can the video be, and what resolution is the output?
Up to 60 seconds, 50 MB and 2,400 frames per clip. The depth map keeps the input's length and frame rate, and its resolution up to 1080p; larger inputs are scaled down to 1080p.
Which video formats work?
MP4 and MOV files from phones, cameras and editing apps. If a clip fails, re-export it as an H.264 MP4. The result is always an H.264 MP4.
How do I use a video depth map in After Effects?
Import the depth MP4, place it above your footage and hide it. Pick it as the depth layer in Camera Lens Blur for depth of field, or use it as a Luma Matte for fog and colour layers. Invert it when an effect expects white to be far.
Can I use it in the free version of DaVinci Resolve?
Yes. The built-in Depth Map effect is a Resolve Studio feature, but a depth map file made here works in the free version: put it on a track above the footage and use it as a mask in Fusion.
How do I use a video depth map in Blender?
Load it as a movie texture and connect it to a Displace modifier on a subdivided plane, or use it as a mask in the compositor. Keep displacement moderate, since 8-bit depth can show steps.
Does it work with ControlNet and ComfyUI?
Yes. The grayscale output is a standard depth control video. Load it with a video loader node and connect it to a depth ControlNet or a video model with depth control, such as Wan or VACE workflows.
Is the depth map 16-bit or metric?
No. You get relative depth (near versus far, not distances in metres) in an 8-bit H.264 MP4. That is enough for mattes, fog, depth of field and ControlNet. For heavy displacement that needs 16-bit or EXR, run the model locally.
Can I use the depth maps in commercial projects?
Yes. The depth model runs on fal's fal-ai/depth-anything-video endpoint, which fal licenses for commercial use, so you can use the depth maps in client work, ads and other paid projects. You still need the rights to the video you upload.
Do I need a GPU or any software?
No. The model runs on fal's servers. You only need a browser and an account.
What happens if processing fails?
Your credits are returned automatically. The usual cause is a file the model cannot read; re-export it as an H.264 MP4 and try again.
Who can see my uploaded video?
Your clip and depth map are never published. Only your account can open them, and the model provider receives the clip through a short-lived link only to process it.
More from this site
Hotel Lobby AI Video Generator
Two photos become a viral Hotel Lobby dance video.
Pricing and credits
Credit packs and plans for every tool on the site.
Turn your next clip into a depth map
Upload an MP4 or MOV, pick a colormap and download the depth video.
Make a depth map