Soniox icon

Soniox

Interact with Soniox Speech-to-Text API

Actions14

Transcription → Transcribe

AI-generated

Summary

Transcribe audio with Soniox by uploading an input binary file or submitting an audio URL, then poll until the job completes and retrieve its transcript. You can select a model and optionally configure transcription context, language handling, translation, speaker diarization, webhooks, polling, output detail, and cleanup.

Inputs

  • Audio Source (required) — Choose Binary Property (default) or URL. For Binary Property, set the required Binary Property name (default data) to match the input item's audio or video binary field. For URL, provide a required audio URL accessible to Soniox servers.
  • Model Name or ID (required) — Select a Soniox transcription model; defaults to stt-async-v5.
  • Additional Fields — Optional settings include a client reference ID; general context as JSON or structured key/value entries; context terms and text; translation term pairs as JSON or structured entries; language identification and hints; speaker diarization; one-way translation with a target language or two-way translation with languages A and B; and a webhook URL with optional authentication header name and value. Do not provide both JSON and structured forms of the same context entries.
  • Options — Configure polling interval (default 5 seconds) and maximum wait (default 300 seconds), local binary file size and duration checks, whether to delete the uploaded file after completion (default true) or transcription after retrieval (default false), and whether to include token-level transcript data (default false). Set local limit values to 0 to disable those checks.

Output shape

One n8n item per input item, containing the Soniox transcription status fields and transcript text; optional transcript details may also be included.

The operation uploads binary audio when that source is selected, creates a transcription, polls immediately and then at the configured interval, and retrieves the transcript on completion. With speaker diarization enabled, token data is grouped into speakers segments; when Include Tokens is enabled, raw tokens are returned. By default, only clean text is returned alongside status fields. The uploaded file and transcription are deleted best effort according to their respective options. The operation errors if transcription fails or exceeds the maximum wait time.

Links

Discussion