GET

/v1/schemas/openrouter/audio/audio%2Ftranscriptions

kind inputactivity audiohash d898f49a…extracted Nov 19, 2026version currentsource upstream spec ↗

Request body

14 properties · 1 required · 0 $defs
propertytypedescriptionconstraints
model*stringThe model to use for transcription.
diarizebooleanLabel each word with the speaker who said it (words[].speaker, words[].speaker_label). Requires response_format "verbose_json" (400 otherwise); word timestamps are included even when timestamp_granularities[] omits "word". Only supported by some providers; 400…
filestringThe audio file to transcribe. The format is derived from the filename extension or the file part content type. Max 25 MB; send larger files as base64 JSON via input_audio, or by URL via source_url. Exactly one of file or source_url is required.format: binary
keyterms[]string[]Domain terms, names, or phrases to bias recognition toward; repeat the part once per term (keyterms=... is also accepted). Only supported by some providers; 400 when the selected model cannot use keyterms.items ≤ 1000
languagestringThe language of the input audio (ISO-639-1).
promptstringContext to guide transcription on providers that support contextual prompting. Takes precedence over provider-specific prompt options, including when empty.
providerstringJSON-encoded provider preferences object, the same shape as the JSON body field: { "zdr": true, "data_collection": "deny", "options": { "<provider-slug>": { ... } } }. Only options for the matched provider are forwarded. Must decode to a JSON object.
response_formatenumThe response format. "json" (default) returns { text, usage }; "verbose_json" additionally returns task, language, duration, and segment-level timestamps (OpenAI-compatible providers only).json | verbose_json
session_idstringA unique identifier for grouping related requests (e.g., a conversation or agent workflow). Used for observability grouping in Broadcast and private logging; never sent to the provider. If provided in both the request body and the x-session-id header, the body…length ≤ 256
source_urlstringPublicly reachable http(s) URL of the audio file, downloaded by the provider directly (no size limit on our side). The format is derived from the URL path extension. Only supported by some providers; exactly one of file or source_url is required.length ≤ 8000 · format: uri
temperaturenumberThe sampling temperature.0 ≤ n ≤ 1
timestamp_granularities[]enum[]Timestamp detail levels to include when response_format is "verbose_json". "word" additionally returns word-level timestamps in the words array.
tracestringJSON-encoded trace metadata object (trace_id, trace_name, span_name, generation_name, parent_span_id and custom keys) attached to the Broadcast trace. Must decode to a JSON object.
userstringA unique identifier representing your end-user. Forwarded to Broadcast and private logging as the end-user id; never sent to the provider.length ≤ 256

Validate a payload

before you spend tokens
shell
$ curl -X POST https://modelschemas.com/v1/validate \
    -d '{"provider":"openrouter","endpointId":"audio/transcriptions","payload":{…}}'
→ {"valid": false, "errors": [{"path": "#", "keyword": "required", …}]}