Meta AI SAM 3 — Segment Anything
Meta AI's SAM 3 is a next-gen vision model for universal detection, segmentation and tracking across images and video. It supports text/image prompts, interactive point/box/mask segmentation and open-vocabulary segmentation. Includes three ready ComfyUI workflows (boxes, points, video tracking).
✓ 100% free — download the workflow and run it locally.Preview
Requirements
- GPU: NVIDIA RTX 30XX/40XX/50XX (FP16; CPU also possible)
- VRAM: 4 GB minimum (more recommended)
- OS: Windows
- Storage: 25 GB+ free
Model downloads
- SAM 3 model (accept the HF agreement first) https://huggingface.co/facebook/sam3
Custom nodes
- ComfyUI Manager https://github.com/ltdrdata/ComfyUI-Manager
- ComfyUI-SAM3 https://github.com/PozzettiAndrea/ComfyUI-SAM3
Setup steps
- Accept the SAM 3 agreement on the HuggingFace model page.
- Generate a HuggingFace access token in your account settings.
- Paste the token into the "Load SAM3 Model" node (or point it at a pre-downloaded model).
- Boxes workflow: left-drag to draw a box, right-click to remove, add a short prompt, run.
- Points workflow: left-click points across the subject, then run.
- Video tracking workflow: track a subject frame-by-frame across a clip.
Notes & tips
- The model auto-downloads on first run once a valid token is provided.
- Short, descriptive prompts improve segmentation quality.
Want the 1-click version?
Skip the manual model downloads and node installs — get a ready-to-run installer, or unlock all 75+ with Local Lab Pro.