Spoken Language Identification

sherpa-onnx supports spoken language identification through the SherpaOnnxCreateSpokenLanguageIdentification() API. It uses Whisper-based models to detect the language of spoken audio.

See also

SherpaOnnxCreateSpokenLanguageIdentification, SherpaOnnxSpokenLanguageIdentificationConfig

SherpaOnnxSpokenLanguageIdentificationConfig config;
memset(&config, 0, sizeof(config));
config.whisper.encoder = "./sherpa-onnx-whisper-tiny/tiny-encoder.int8.onnx";
config.whisper.decoder = "./sherpa-onnx-whisper-tiny/tiny-decoder.int8.onnx";
config.num_threads = 1;
config.provider = "cpu";
 
const SherpaOnnxSpokenLanguageIdentification *slid =
    SherpaOnnxCreateSpokenLanguageIdentification(&config);
 
SherpaOnnxOfflineStream *stream =
    SherpaOnnxSpokenLanguageIdentificationCreateOfflineStream(slid);
SherpaOnnxAcceptWaveformOffline(stream, wave->sample_rate,
                                wave->samples, wave->num_samples);
 
const SherpaOnnxSpokenLanguageIdentificationResult *result =
    SherpaOnnxSpokenLanguageIdentificationCompute(slid, stream);
printf("Detected language: %s\n", result->lang);
 
SherpaOnnxDestroySpokenLanguageIdentificationResult(result);
SherpaOnnxDestroyOfflineStream(stream);
SherpaOnnxDestroySpokenLanguageIdentification(slid);

Model package: sherpa-onnx-whisper-tiny

Example source: spoken-language-identification-c-api.c