GET

/v1/schemas/fal/audio/fal-ai/sam-audio/span-separate

Accept:

kind inputactivity audiohash d0607ebf…extracted Jul 6, 2026version currentsource upstream spec ↗

field	recorded source
schema	Source linked above. Source trace `134e9fc601c38ed6e06a4598efb599d9e2576736b397150f13909e6ecea70f29` { "fetchedAt": 1783330412, "extractorVersion": "1", "verifiedAt": null }

Request body

10 properties · 2 required · 1 $defs

property	type	description	constraints
audio_url*	string	URL of the audio file to process.
spans*	AudioTimeSpan[]	Time spans where the target sound occurs which should be isolated.
output_format	enum	Output audio format.	wav \| mp3 · default: "wav"
chunk_overlap	number	Overlap duration (in seconds) between chunks for crossfade blending.	0 ≤ n ≤ 30 · default: 5
prompt	string	Text prompt describing the sound to isolate. Optional but recommended - helps the model identify what type of sound to extract from the span.	nullable
acceleration	enum	The acceleration level to use.	fast \| balanced \| quality · default: "balanced"
reranking_candidates	integer	Number of candidates to generate and rank. Higher improves quality but increases latency and cost. Requires text prompt; ignored for span-only separation.	1 ≤ n ≤ 7 · default: 1
max_chunk_duration	number	Maximum audio duration (in seconds) to process in a single pass. Longer audio will be chunked with overlap and blended.	10 ≤ n ≤ 60 · default: 60
trim_to_span	boolean	Trim output audio to only include the specified span time range. If False, returns the full audio length with the target sound isolated throughout.	default: false
use_sound_activity_ranking	boolean	Use sound activity detection to rank reranking candidates based on how well each candidate's non-silent regions match the provided spans. Enables effective reranking even without a text prompt (span-only separation). Requires reranking_candidates > 1.	default: false

Types

1 $defs

AudioTimeSpanobject

A time span indicating where the target sound occurs.

include	boolean	Whether to include (True) or exclude (False) sounds in this span	default: true
start*	number	Start time of the span in seconds	n ≥ 0
end*	number	End time of the span in seconds	n ≥ 0

Validate a payload

before you spend tokens

shell

$ curl -X POST https://modelschemas.com/v1/validate \
    -d '{"provider":"fal","endpointId":"fal-ai/sam-audio/span-separate","payload":{…}}'
→ {"valid": false, "errors": [{"path": "#", "keyword": "required", …}]}