depth-map comparison free-tools
The best free depth map generators in 2026, compared (including where Gifsy loses)
Free depth map generators compared on what matters: which model, whether your photo is uploaded, sign-up, limits, output bit depth and resolution, with the cases where another tool is the better pick.
Aman Jha·September 25, 2026·8 min read
A depth map generator turns one photo into a grey image where brightness means distance. They all do that. Where they differ is the model underneath, whether your photo leaves your machine, and what file you get back. Nine free options, compared on what each tool's own page says on 25 September 2026. We make one of them, so Gifsy gets a row like everyone else, plus a section on where it loses.
The short answer
For the best raw map, use the official Depth Anything V2 demo on Hugging Face: it runs the Large model on a server and gives a 16-bit file. For a private photo, use an in-browser tool, Gifsy for a 16-bit PNG with no sign-up or kordu.tools for colour maps. DepthR is the free pick for an STL relief.
Side by side
A green tick means the answer is good for you, a red cross means it costs you something. Plain text means it depends, or the tool's page doesn't say. Scroll sideways on a phone.
The nine tools
Depth Anything V2 demo (Hugging Face)
This is the model's own demo, and the code behind it is public. It runs the Large version of Depth Anything V2 on Hugging Face's shared GPUs and hands back three things: a coloured slider view, a grayscale PNG and a “16-bit raw output (can be considered as disparity)”. Large is a bigger model than the Small one most browser tools use, so it is the reference for what the method can do. Two things to know. Your photo goes to a server. And the Depth Anything V2 repo says Base, Large and Giant are under CC-BY-NC-4.0, a non-commercial licence, while Small is Apache-2.0.
Pick it for: the best raw map, when the photo isn't private.
Gifsy
Ours. Gifsy's depth map generator runs Depth Anything V2 Small in the tab with ONNX Runtime Web. The 50 MB model downloads once and stays cached; after that nothing leaves your machine, and there is no account and no limit. You get a grayscale PNG, white for near, at your photo's size up to 2048 px, in 8-bit or 16-bit. Two switches: Invert, for apps that want it the other way round, and Sharpen edges, a guided filter that snaps depth edges to the photo's own edges.
We built it because we didn't want to store anyone's photos, and a depth map is small enough to make on the visitor's own laptop. How that works, including the 16-bit PNG encoder, is in running Depth Anything V2 in the browser.
Pick it for: a private photo, a 16-bit file, and no sign-up.
kordu.tools
The closest thing to Gifsy here, and it deserves the credit. kordu.tools also runs Depth-Anything Small in your browser: “Your images never leave your device.” The model is about 50 MB, the first run takes 15 to 60 seconds, and inputs go up to 20 MB. What it has that Gifsy doesn't is colour: Grayscale, Viridis, Plasma and Inferno. Its page doesn't state bit depth or output size.
Pick it for: a coloured depth map without uploading.
teamzlab
teamzlab's AI depth map is another in-browser Depth Anything tool (“No image data ever leaves your device”). It shows a viridis colour view and a grayscale one where “white is near and black is far”. File size limits, resolution and bit depth aren't stated.
Pick it for: a quick private look at a photo's depth, in colour.
Upsampler
Upsampler runs on its own servers with no sign-up. The free tier is a “free daily allowance of GPU minutes”, and images “are processed on-the-fly and are not stored permanently”. It exports JPG, PNG or WebP, says plainly “We do not add any watermarks”, and allows commercial use. The model isn't named beyond a “monocular depth model”.
Pick it for: a server-side map with no account, when uploading is fine.
DepthR
DepthR is aimed at relief and 3D printing: “Download the finished result as STL or PNG Depth Map”, with an upscale option. Free use is 3 depth maps a day. Beyond that it's a paid licence, from $26.90 for a month on its pricing page when we checked. It doesn't name its model or say where processing happens.
Pick it for: going straight from photo to STL.
Artificial Studio
Artificial Studio is a server tool inside a larger AI app, priced in credits: $12 for 1,000, one credit per depth map. Its selling point for developers is an API. For video it suggests extracting frames and running them one at a time. The model, bit depth and resolution aren't stated.
Pick it for: batch jobs through an API.
Dreamina
CapCut's Dreamina makes depth maps through its AI image tool, and you have to log in first. What it offers that nobody else here states is size: “High (2K)” or “Ultra (4K)”. Its page names the Seedream 4.0 model, which is an image generator rather than a dedicated depth estimator, and it doesn't state the file format or bit depth.
Pick it for: a 2K or 4K map, if you already use CapCut's tools.
Vector Witch
Vector Witch runs in the browser (“no uploads needed”) and is “optimized for laser engraving”, with inputs up to 25 MB. Read its description closely, though: “Luminance-based grayscale mapping with adjustable depth levels”. That turns brightness into height, so a dark sweater and a pale wall land at different heights because of their colour, not their shape. For some engraving jobs that's what you want. For the shape of a scene, it isn't.
Pick it for: brightness-to-height engraving, where you know that's the effect.
Which one for which job
- The best map you can get for free: the Hugging Face demo, if the photo can go to a server and the use is non-commercial.
- A private photo: Gifsy, kordu.tools or teamzlab. All three run a depth model in your browser.
- Displacement in Blender: a 16-bit file, so Gifsy or the Hugging Face demo. Our Blender depth map guide covers the Displace settings.
- Laser engraving: a depth model for shape, 16-bit if your software reads it. See the depth map for laser engraving page for how LightBurn reads the grey.
- An STL for printing or CNC: DepthR.
- A coloured map for a slide or a paper: kordu.tools or teamzlab.
- 2K or 4K output: Dreamina.
Where Gifsy falls short
A comparison written by one of the contestants should say where it loses. So:
- The model sees 518 px on the long edge, or 770 on a machine with 8 GB of RAM and 8 cores. Anything bigger is upscaled. A 2048 px download is not 2048 px of real detail.
- Small, not Large. The Hugging Face demo runs a bigger model.
- Grey only. No colour maps; kordu.tools and teamzlab have them.
- No STL or mesh export. DepthR does that.
- No video. It's one photo at a time.
- The first run downloads about 50 MB, which is slow on a phone connection.
- Relative depth, like every tool here. It knows the order of things, not their distance in metres.
FAQ
What is the best free depth map generator?
For raw quality, the official Depth Anything V2 demo on Hugging Face, because it runs the Large model on a server and gives you a 16-bit file. For a private photo, an in-browser tool: Gifsy (16-bit PNG, no sign-up, no limit) or kordu.tools (colour maps). For an STL relief, DepthR, which allows 3 free maps a day.
Which depth map generators don't upload my photo?
Gifsy, kordu.tools, teamzlab and Vector Witch all say the work happens in your browser. Upsampler, Artificial Studio and the Hugging Face demo process the image on a server; Upsampler says images are not stored permanently. DepthR and Dreamina don't say where processing happens.
Is there a free depth map generator with no sign-up?
Yes. Gifsy needs no account and has no daily limit. Upsampler needs no sign-up but caps free use with a daily allowance of GPU minutes. Dreamina needs a login, DepthR allows 3 free maps a day with paid plans beyond that, and Artificial Studio works on credits.
Which free depth map generator gives a 16-bit file?
Gifsy's download has a 16-bit PNG option, and the official Hugging Face demo returns a "16-bit raw output". None of the other tools here states a bit depth. 16-bit matters when the map drives displacement or a laser, because 8 bits gives only 256 steps.
Can I get a depth map as an STL for 3D printing or CNC?
Of the tools compared here, DepthR says you can download the result as STL or PNG. Gifsy, kordu.tools, Upsampler and the others give an image only, which you would then turn into a mesh in Blender or a CAM tool.
Which one is best for laser engraving?
Vector Witch markets itself for engraving but describes luminance-based mapping, which follows brightness rather than distance. A depth model (Gifsy, kordu, the Hugging Face demo) follows the shape of the scene. Gifsy has a laser page that covers which way round LightBurn reads the grey.
Are these depth maps in metres?
No. The tools here built on Depth Anything estimate relative depth from a single photo, meaning which things are nearer and which are farther. The Hugging Face demo labels its 16-bit output as disparity. None of them gives distances you could measure.
How this was compared
Every fact about another tool comes from that tool's own page (its tool page, pricing page or, for the Hugging Face demo, its public app code), read on 25 September 2026. We didn't run the same photo through each tool; that's a hands-on test we haven't done yet, which is why this page says compared rather than tested. Where a page doesn't say, the table says “not stated” rather than guessing. Gifsy's numbers come from its own source code. We make Gifsy; the “pick it for” line on every tool and the section on where Gifsy loses are how we try to keep that honest.
Update log
- 25 September 2026: first published. Nine tools. Sculptok left out because its page states no free tier, formats or limits we could quote.
Gifsy's depth map tool is free and runs in your browser. If you want the photo itself to move in 3D on a website, that's the 3D parallax photo maker, a separate tool with its own free tier.