Try out this capability in Bria's sandbox

Description Initiates an asynchronous background removal job for a video.

Need real-time / low-latency background removal? See Streaming Background Removal for the WebSocket-based variant.

Response behavior:

• Returns HTTP 202 with request_id and status_url

• Check job status by polling status_url or by calling the status endpoint with the request_id until a terminal status is returned

Supported input containers:

• .mp4, .mov, .webm, .avi, .gif

Supported input codecs:

• H.264, H.265 (HEVC), VP9, AV1, PhotoJPEG

Input attributes preserved in output:

• Aspect Ratio and resolution (unless auto_zoom is enabled - see Auto Zoom below)

• Frame Rate

• Audio if present

Limits:

• Max input duration: 60 seconds

• Resolutions: Input resolution up to 16000x16000 (16K). Inputs larger than this return 413 Payload Too Large. Output resolution matches the input resolution. When auto_zoom is enabled, the output is cropped to the foreground subject and may be smaller than the input.

Background and transparency control:

• background_color must be one of the predefined string values listed below. Hex values are not supported. Default is Transparent.

• If background_color is Transparent and the selected output preset cannot carry alpha, the request is not rejected. background_color falls back to Black and the completed result carries a warning field naming the preset, the colour used, and the presets that do support transparency. Poll the status endpoint and check result.warning if you need to detect this.

Output transparency support by preset:

• Alpha supported: webm_vp9, mov_proresks, mkv_vp9

• Alpha not supported: mp4_h264, mp4_h265, mov_h265, mkv_h264, mkv_h265, gif

Auto Zoom:

• Set auto_zoom to true to automatically crop the output video to the foreground subject, creating a zoom-in effect.

• The crop is calculated once for the entire video - the smallest rectangle that contains the subject across all frames - so the framing is stable throughout, with no per-frame movement.

• If the subject moves across a large portion of the frame, the crop is minimal and the output may match the input framing.

• When cropping is applied, the output resolution and aspect ratio differ from the input.

• Enabling this option increases processing time.