ailia_speech/axip.ailia_speech/AiliaSpeech

AiliaSpeech

class AiliaSpeech(envId: Int = -1, numThread: Int = 0, memoryMode: Int = 11, task: Int = AILIA_SPEECH_TASK_TRANSCRIBE, flags: Int = AILIA_SPEECH_FLAG_NONE)

Speech recognition for audio waveforms.

This class provides speech to text conversion (Whisper and SenseVoice), voice activity detection, speaker diarization and post processing such as translation via the ailia Speech native library.

Feed the audio waveform with pushInputData, notify the end of the input with finalizeInputData, then call transcribe while getBuffered returns 1 and get the results with getText. getComplete returns 1 when all of the input has been processed.

Parameters

envId

Environment id of the ailia execution. -1 (ENVIRONMENT_ID_AUTO) selects the environment automatically.

numThread

Number of threads. 0 (MULTITHREAD_AUTO) means the system's logical processor count.

memoryMode

Memory management mode of the ailia execution. The default value 11 reduces the memory consumption.

task

Task of the recognition, one of the AILIA_SPEECH_TASK_* constants.

flags

Flags of the recognition, one of the AILIA_SPEECH_FLAG_* constants.

Constructors

constructor(envId: Int = -1, numThread: Int = 0, memoryMode: Int = 11, task: Int = AILIA_SPEECH_TASK_TRANSCRIBE, flags: Int = AILIA_SPEECH_FLAG_NONE)

Creates an ailia Speech instance.

Types

object Companion

Constants of ailia Speech.

Functions

fun close()

Destroys the ailia Speech instance.

finalizeInputData

fun finalizeInputData(): Int

Notifies the end of the audio input.

fun getBuffered(): Int

Determines if enough audio data has been fed to perform the transcription.

fun getComplete(): Int

Determines if all of the audio data has been processed.

fun getErrorDetail(): String

Returns the details of the last error.

fun getText(idx: Int): AiliaSpeechText

Returns the recognized text fragment.

fun getTextCount(): Int

Returns the number of the recognized text fragments.

openDiarization

fun openDiarization(segmentationPath: String, embeddingPath: String, diarizationType: Int): Int

Opens the speaker diarization model files.

fun openDictionary(dictionaryPath: String, dictionaryType: Int): Int

Opens a dictionary file for correcting the recognition results.

fun openModel(encoderPath: String, decoderPath: String, modelType: Int): Int

Opens the speech recognition model files.

openPostProcess

fun openPostProcess(encoderPath: String, decoderPath: String, sourcePath: String, targetPath: String, prefix: String, postProcessType: Int): Int

Opens the post process model files for error correction or translation of the recognition results.

fun openVad(vadPath: String, vadType: Int): Int

Opens the voice activity detection (VAD) model file.

fun postProcess(): Int

Performs the post process (error correction or translation) of the recognition results.

fun pushInputData(src: FloatArray, channels: Int, samples: Int, samplingRate: Int): Int

Pushes the audio waveform to the queue.

resetTranscribeState

fun resetTranscribeState(): Int

Resets the transcribe state to process a new audio input.

fun setConstraint(constraint: String, type: Int): Int

Sets the constraint of the recognition output.

setIntermediateCallback

fun setIntermediateCallback(callback: IntermediateCallback): Int

Sets the callback for receiving intermediate recognition results.

fun setLanguage(language: String): Int

Sets the language of the speech recognition.

fun setPrompt(prompt: String): Int

Sets the prompt text given to the model.

setSilentThreshold

fun setSilentThreshold(silentThreshold: Float, speechSec: Float, noSpeechSec: Float): Int

Sets the silent threshold of the voice activity detection.

fun transcribe(): Int

Performs the speech recognition of the buffered audio data.