FishAudioSpeechToText - ComfyUI Built-in Node Documentation - ComfyUI

This node transcribes audio into text using the Fish Audio speech-to-text service. It automatically detects the language of the audio and can optionally return word-level timestamped segments as JSON.

Inputs

Parameter	Description	Data Type	Required	Range
`audio`	Audio to transcribe.	AUDIO	Yes	—
`language`	ISO 639-1 language hint (e.g. ‘en’, ‘zh’). The language is auto-detected regardless. Default: "" (empty string).	STRING	No	Any ISO 639-1 language code, e.g. `en`, `zh`; empty string for auto-detection
`precise_timestamps`	Return word-level timestamped segments. Default: false.	BOOLEAN	No	true or false

Note: The language parameter is only a hint — the language is always auto-detected from the audio. When precise_timestamps is false (the default), word-level timestamps are not returned; when true, the output segments include word-level timestamps.

Outputs

Output Name	Description	Data Type
`text`	The transcribed text.	STRING
`language_code`	The ISO 639-1 language code detected for the audio.	STRING
`segments_json`	JSON string containing the transcription segments. Includes word-level timestamps when `precise_timestamps` is enabled.	STRING

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

Source fingerprint (SHA-256): eaf1c9a9d2b90ec962a408615cc417b552864354c3f272144b8e239b23961920