The Chinese result is wrong, but whisper.cpp is correct.

[j20cc]

use whisper.rn get

use whisper.cpp get

here is my code, they use same model and input audio

const whisper = async () => {
    const path = ReactNativeBlobUtil.fs.dirs.DocumentDir;
    const whisperContext = await initWhisper({
      filePath: `file://${path}/ggml-small.bin`,
    });

    const sampleFilePath = `file://${path}/1.wav`;
    console.log('sampleFilePath:', sampleFilePath);

    console.log('Start transcribing...');
    const startTime = Date.now();
    const {stop, promise} = whisperContext.transcribe(sampleFilePath, {
      language: 'auto',
      prompt: '以下是普通话',
      onNewSegments: segments => {
        console.log('New segments:', segments);
      },
    });
    const {result, segments} = await promise;
    console.log('result:', result);
  };

the language option i also tried 'chinese', not work correct too, whisper.cpp result is correct