SAM 3 Image
AI generationMetaSingle-frame segmentation - extract masks from images with text prompts.
- Runs
- On our servers
- Typical time
- 5-30s
- Inputs
- 2
- Outputs
- 1
Inputs
| Port | Type | ID |
|---|---|---|
| Prompt | text | prompt-in |
| Image | image | image-in |
Outputs
| Port | Type | ID |
|---|---|---|
| Image | image | image-out |
Settings
| Setting | Type | Default | Options |
|---|---|---|---|
| Output ModeCutout returns the subject on a transparent background. Alpha Mask returns a black-and-white matte to use in Merge, Keymix, or Copy. Ignored when Multiple Masks is on (each object comes back as its own matte). | select | true | Cutout (object only), Alpha Mask (B&W matte) |
| Multiple MasksSingle: one combined matte. Multiple: a separate B&W matte per comma-separated subject and per instance ("wheel" returns each wheel), up to Max Masks - one Read per matte in Nuke, selectable outputs in Studio. Mattes load into the red channel in Nuke: aim a mask picker at .red or Shuffle red to alpha. Overrides Output Mode. | select | false | Single mask, Multiple masks |
| Max MasksCap on how many instance mattes come back when Multiple Masks is on (provider range 1-32). | slider | 3 | 1 to 32 |
Prompting
Name your subject - or comma-separate several ("giraffe, truck") for a matte each. Avoid "A and B" phrasing: that merges into one...
Maximum prompt length: 500 characters.
What this connects into (24)
Computed from the port compatibility rules, so it stays correct as nodes change.
- Flux LoRA
- Nano Banana 2
- Nano Banana Pro
- GPT Image 2
- Gemini Omni
- Seedance 2.0
- Grok Imagine 1.5 I2V
- Kling 3.0 I2V
- Kling 3.0 Pro I2V
- Kling O3 I2V
- Kling O1 FLF2V
- WAN 2.7 I2V
- HappyHorse 1.0 I2V
- WAN Pro I2V
- WAN Effects
- Seedance 2.0 I2V
- Vidu Q1 FLF2V
- Kling O3 Reference
- Kling O1 Reference
- Seedance 2.0 Reference
- Hyper3D Rodin
- Hunyuan3D v3
- Trellis
- SHARP Gaussian
Node ID sam3-image. Open it in Studio with Tab, then type the name.