AI Depth Map Generator
Commercial use OK
380+ models
No watermark
No sign-up needed
Model:
+ GPT-5, Claude, Gemini
Generate a depth map from any photo. Useful as a ControlNet conditioning input, for parallax / 3D effects, or as a starting point for AR/VR scene reconstruction. Powered by Marigold (newer, sharper) or fal Depth (faster).
Advanced options
Резултат
Tokens running low.
Get More Tokens
❤️ Love Free.ai? Tell your friends!
Sign up to get a referral link and earn 25,000 tokens per friend.
Сакаш уште?
Sign up free for 10,000 tokens
Слободно потпиши се
Го обработувам вашето барање...
Generate a depth map from any photo with AI. Free Marigold (sharp, self-hosted) or premium fast depth — outputs ControlNet-ready depth PNG for parallax, AR/VR, 3D scene reconstruction.
Како да се користи AI Depth Map Generator
1
Внесете го вашиот влез
Внесете текст, внесете датотека или опишете што сакате. Нема потреба од сметка.
2
Кликнете на генерирање
Нашиот АИ го процесира вашето барање во секунди користејќи ги најдобрите модели на отворен извор.
3
Симни & споделување
Бесплатно за лична и комерцијална употреба.
Use this tool via API
Automate this tool from your own code. OpenAI-compatible REST endpoint, Bearer-token auth, no extra SDK required. Token costs match the web interface.
curl -X POST https://api.free.ai/v1/image/generate/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"prompt": "A sunset over mountains", "model": "sdxl"}'Соодветни слободни AI алатки
AI Depth Map Generator — FAQ
Upload any photo, get a depth map back — a colorized image where each pixel encodes how far that point is from the camera. Use it as a ControlNet input for depth-conditioned image generation, as a depth pass for parallax / 3D effects in After Effects or Blender, or as a starting point for AR / VR scene reconstruction. Two models: free Marigold (sharp, slower) or premium fast depth (less detail, faster on huge inputs).
Yes. Marigold runs on our self-hosted A100 GPUs — pulls from your daily token pool (2,500/day anonymous, 10,000/day signed up). Each depth map is ~1,500 tokens. Premium fast-depth costs the same baseline but can scale better on 4K+ inputs.
Marigold (default) gives the sharpest, most accurate depth — fine-tuned from Stable Diffusion, picks up subtle depth in plants / hair / cloth folds. Premium is quicker on >2K inputs and more reliable on extreme wide-angle shots. For ControlNet conditioning or 3D reconstruction, default to Marigold. For batch processing thousands of images, premium scales better.
A 512×512 RGB PNG with a Spectral colormap — bright yellow / orange = near, deep blue / purple = far. The raw depth values are normalized 0-1 across the image. If you need 16-bit grayscale depth (for ControlNet or Blender displacement maps), use the API and pass `format=16bit_png`.
Yes — Marigold output is the standard ControlNet depth-conditioning format. Use it in Stable Diffusion / FLUX with the depth ControlNet for camera-aware re-rendering: change subject, lighting, style while preserving 3D structure. Free.ai chat can chain it for you — say "depth map this, then re-render with FLUX in the style of Studio Ghibli".
Marigold is the leading open-source monocular depth estimator (CVPR 2024 best paper). Within ~3-5% of ground truth on indoor benchmarks (NYU Depth v2), within 5-8% outdoors (KITTI). For metric depth (real-world meters), pair it with a calibrated camera. For relative depth (this is closer than that), accuracy is excellent.
JPG, PNG, WebP up to 10MB. Marigold runs at 768×768 internally regardless of input size and resamples back. Output is always 512×512 RGB. For higher-resolution depth, use the premium fast option which scales linearly.
Yes — depth maps are the standard input for After Effects 2.5D parallax (Camera Mapper plugin), DaVinci Resolve depth-based blur, and Blender displacement modeling. Generate depth, import as an alpha-matte sequence, key your camera moves to it. Many photographers use this to fake portrait-mode bokeh on phones without dual cameras.
Marigold (2024) outperforms MiDaS v3 (2022) and Intel DPT on every public benchmark, especially on fine details (hair, foliage, transparency). Paid depth APIs (Replicate, fal) charge $0.005-0.02 per image — we run the same Marigold weights free. Quality difference: zero.
Yes — POST to /v1/image/depth-map/ with multipart image. Or use /batch/ to drop a folder and get a ZIP back. Useful for processing entire video frame sequences for depth-aware effects.
Yes — POST multipart `image` to /v1/image/depth-map/ on api.free.ai. Bearer auth with your API key. 10,000 tokens/month free, paid tiers above. Returns JSON with `image_url` to the depth PNG. /api/ has the curl example.
Drop into Photoshop as a layer-mask source, ControlNet condition for re-rendering in /image/edit/, displacement map in Blender, parallax animator in After Effects, or 2.5D dolly-zoom video creator (LeiaPix-style). Chain with /image/edit/ for depth-conditioned generation in one tool.
How would you rate this tool?