transcribe()

transcribe() is a suspend extension function on RunAnywhere. It takes raw audio bytes and returns an STTOutput, not a bare String.

import ai.runanywhere.proto.v1.ModelCategory
import ai.runanywhere.proto.v1.STTLanguage
import com.runanywhere.sdk.public.RunAnywhere
import com.runanywhere.sdk.public.extensions.loadModel
import com.runanywhere.sdk.public.extensions.transcribe
import com.runanywhere.sdk.public.types.RAModelLoadRequest
import com.runanywhere.sdk.public.types.RASTTOptions

RunAnywhere.loadModel(
    RAModelLoadRequest(
        model_id = "sherpa-onnx-whisper-tiny.en",
        category = ModelCategory.MODEL_CATEGORY_SPEECH_RECOGNITION,
    ),
)

val output = RunAnywhere.transcribe(
    audio = audioBytes,
    options = RASTTOptions(
        language = STTLanguage.STT_LANGUAGE_EN,
        enable_punctuation = true,
    ),
)

println(output.text)
println(output.confidence)

Signature:

suspend fun RunAnywhere.transcribe(
    audio: ByteArray,
    options: RASTTOptions = RASTTOptions.defaults(),
): RASTTOutput

There is no transcribeWithOptions; transcribe() always takes options and always returns the detailed output. RASTTOptions is a typealias for the generated ai.runanywhere.proto.v1.STTOptions, and RASTTOutput for STTOutput. Audio must be 16 kHz mono 16-bit PCM.

Model loading

There is one load entry point for every modality. Category is optional and only needed to disambiguate a model registered under more than one:

RunAnywhere.loadModel(RAModelLoadRequest(model_id = "sherpa-onnx-whisper-tiny.en"))

loadModel returns a ModelLoadResult with a success flag rather than throwing:

val result = RunAnywhere.loadModel(request)
if (!result.success) showError(result.error_message)

To see what is currently loaded for speech recognition:

import ai.runanywhere.proto.v1.CurrentModelRequest
import com.runanywhere.sdk.public.extensions.currentModel

val snapshot = RunAnywhere.currentModel(
    CurrentModelRequest(category = ModelCategory.MODEL_CATEGORY_SPEECH_RECOGNITION),
)
if (snapshot.found) println(snapshot.model_id)

STTOutput

Field	Type	Description
`text`	`String`	Transcribed text
`language`	`STTLanguage`	Language enum
`language_code`	`String?`	Free-form detected tag, preserves regional variants
`confidence`	`Float`	0.0 to 1.0
`words`	`List<WordTimestamp>`	Word-level timing
`alternatives`	`List<TranscriptionAlternative>`	Competing hypotheses
`metadata`	`TranscriptionMetadata?`	Processing metrics
`duration_ms`	`Long`	Audio duration
`timestamp_ms`	`Long`	Wall clock, ms since epoch
`speaker_ids`	`List<String>`	Populated when diarization is on
`segment_index`	`Int`	Segment index for long or streaming transcription
`error_message`	`String?`	Terminal error detail
`error_code`	`Int`	`0` when unset

WordTimestamp carries word, start_ms, end_ms, confidence, and speaker_id. All timings are milliseconds, as Long.

output.words.forEach { word ->
    println("[${word.start_ms}ms - ${word.end_ms}ms] ${word.word}")
}

TranscriptionMetadata carries model_id, processing_time_ms, audio_length_ms, and real_time_factor. A real-time factor below 1.0 means transcription ran faster than the audio plays.

val audioMs = output.duration_ms.takeIf { it > 0 }
    ?: output.metadata?.audio_length_ms ?: 0L
val processingMs = output.metadata?.processing_time_ms ?: 0L

Both are nullable-safe reads because metadata is an optional message and either field can be 0 when the backend does not report it.

Transcribing a file

suspend fun transcribeAudioFile(uri: Uri): String {
    val audio = contentResolver.openInputStream(uri)?.readBytes()
        ?: throw IllegalArgumentException("Cannot read audio file")
    return RunAnywhere.transcribe(audio).text
}

The file has to contain 16 kHz mono 16-bit PCM samples. Decode or resample anything else before calling.

Errors

transcribe() throws SDKException when the SDK is not initialized or when no speech-recognition model is loaded. See Error handling.

Streaming STT

Live microphone transcription

STT options

Every field on STTOptions

Getting Started

Swift SDK

Kotlin SDK

React Native SDK

Flutter SDK

Web SDK

Vibe Coding

Model loading

STTOutput

Transcribing a file

Errors

Streaming STT

STT options

​Model loading

​STTOutput

​Transcribing a file

​Errors

​Related

Streaming STT

STT options

Model loading

STTOutput

Transcribing a file

Errors

Related