All Models

WaveSpeed logo

SAM3 Video API

SAM3 Video segments subjects selected by text and tracks them across footage, producing masks or masked video for compositing and localized editing.

Try in Studio
More options (1)
Generate

Pricing depends on the selected inputs. See the rates below.

Supported modes

Edit

Edit uses prompt and video. Adjust the model options to prepare a request for this mode.

video.edit

Pricing

Video editing

Based on verified input duration.

OfferingPrice (USD)
All listed options$0.01 / input second
  • Apply Mask: Off, On (same price).
  • Based on the video input; billing steps apply in this order:
  • Round input duration up to a whole second.
  • Cap billed duration at 600 seconds.
  • Bill at least 3 seconds.
  • Round the total charge up to the next $0.01.

Edit options

OptionDetailsDefault
Prompt *
prompt
Text prompt for segmentation. Use commas to track multiple objects (e.g., 'person, cloth').โ€”
Video *
video
A public URL, a data URI, or an asset id.โ€”
Apply mask
apply_mask
Whether to apply mask to videotrue

SAM3 Video

SAM3 Video turns a written subject description into a selection that follows the subject through footage. It can return a mask-style video or apply the selection to the source video, providing material for further editing.

The video supplies the scene and movement, while the text identifies what to segment. Selections can include people, clothing, animals or other visible objects, including multiple targets in the same shot. Tracking carries those selections across frames so they can be used throughout a moving sequence.

What it's good for

Editors can prepare a performer for compositing, separate foreground elements for localized treatment, or identify unwanted background objects for cleanup. The masks can also guide removal and replacement work in downstream editing tools. Mask output suits workflows that need a separate selection layer; masked footage provides an isolated subject to carry into a composite.

Run SAM3 Video on Protoface

Use SAM3 Video in Protoface Studio without writing code. Choose your inputs and generation settings, then generate with affordable, usage-based pricing. For your own AI app or workflow, use the Protoface API to submit generations and retrieve their outputs.

Model publisher

About WaveSpeed

WaveSpeed is an AI inference platform offering third-party generative models through developer APIs and creator applications. Its image and video workflows sit alongside WaveSpeed-branded media tools, including Video Face Swap, which takes source footage and a reference portrait to replace a selected face in a video.

  • Media inference
  • Creative tools
  • Video face swap

Related models

Unified API

One API for generative media.

Protoface serves popular frontier AI models through one API.

  • minimax
  • minimax
  • z.ai
  • z.ai
  • z.ai
  • z.ai
  • z.ai
  • z.ai
  • z.ai

Unified API

One API for generative media.

Protoface serves popular frontier AI models through one API.

  • minimax
  • minimax
  • z.ai
  • z.ai
  • z.ai
  • z.ai
  • z.ai
  • z.ai
  • z.ai

Unified API

One API for generative media.

Protoface serves popular frontier AI models through one API.

  • minimax
  • minimax
  • z.ai
  • z.ai
  • z.ai
  • z.ai
  • z.ai
  • z.ai
  • z.ai