Skip to main content
Interface for configuring audio transcription behavior. Used with transcribe(), transcribeData(), and transcribeRealtime() methods.

​
Properties

​
language
string
default:"auto"
Spoken language code (e.g., ‘en’, ‘es’, ‘fr’). Set to ‘auto’ for automatic language detection.
​
translate
boolean
default:"false"
Translate from source language to English. When enabled, the transcription will be translated to English regardless of the input language.
​
maxThreads
number
default:"2 or 4"
Number of threads to use during computation. Default is 2 for 4-core devices, 4 for devices with more cores.
​
nProcessors
number
default:"1"
Number of processors to use for parallel processing with whisper_full_parallel. Set to 1 to use whisper_full instead.
​
maxContext
number
Maximum number of text context tokens to store. Controls how much previous transcription context is retained.
​
maxLen
number
Maximum segment length in characters. Limits the length of individual transcription segments.
​
tokenTimestamps
boolean
Enable token-level timestamps. When enabled, provides more granular timestamp information for each token.
​
tdrzEnable
boolean
Enable tinydiarize speaker diarization. Requires a tdrz model to be loaded.
​
wordThold
number
Word timestamp probability threshold. Controls the confidence threshold for word-level timestamps.
​
offset
number
Time offset in milliseconds. Specifies where to start transcription in the audio file.
​
duration
number
Duration of audio to process in milliseconds. If set, only processes the specified duration from the offset.
​
temperature
number
Initial decoding temperature. Controls the randomness in the decoding process. Higher values increase randomness.
​
temperatureInc
number
Temperature increment value. Used when adjusting temperature during decoding.
​
beamSize
number
Beam size for beam search decoding. Larger values can improve accuracy but increase computation time.
​
bestOf
number
Number of best candidates to keep during decoding. Higher values may improve accuracy at the cost of performance.
​
prompt
string
Initial prompt text to guide transcription. Provides context to improve transcription accuracy for specific terminology or style.

​
Usage Example

import { initWhisper } from 'whisper.rn'

const context = await initWhisper({
  filePath: 'path/to/model.bin'
})

const options: TranscribeOptions = {
  language: 'en',
  translate: false,
  maxThreads: 4,
  tokenTimestamps: true,
  prompt: 'Technical discussion about React Native'
}

const { promise } = context.transcribe('path/to/audio.wav', options)
const result = await promise

Build docs developers (and LLMs) love

Get started for freeTalk to us