go-whisper-api/transcode/decode.go

package transcode

import (
    "fmt"
    "io"
    "os"
    "path/filepath"
    "strings"

    "github.com/gopxl/beep"
    "github.com/gopxl/beep/flac"
    "github.com/gopxl/beep/mp3"
    beepwav "github.com/gopxl/beep/wav"
)

var probeFormats = []string{
    ".wav", ".wave", ".mp3", ".flac", ".ogg", ".opus",
    ".m4a", ".m4b", ".mp4", ".mov", ".m4v", ".3gp", ".3g2", ".aac",
}

func supportedFormatsMessage() string {
    return strings.Join(probeFormats, ", ")
}

func resolveFormat(path string) (string, error) {
    ext := strings.ToLower(filepath.Ext(path))
    if ext != "" {
        return ext, nil
    }
    f, err := os.Open(path)
    if err != nil {
        return "", err
    }
    defer f.Close()
    ext = sniffFormat(f)
    if ext == "" {
        return "", fmt.Errorf("could not detect audio format (supported: %s)", supportedFormatsMessage())
    }
    return ext, nil
}

func openDecoder(path string) (beep.Streamer, beep.Format, io.Closer, error) {
    ext, err := resolveFormat(path)
    if err != nil {
        return nil, beep.Format{}, nil, err
    }
    streamer, format, closer, err := decodeByExt(path, ext)
    if err == nil {
        return streamer, format, closer, nil
    }
    for _, try := range probeFormats {
        if try == ext {
            continue
        }
        streamer, format, closer, tryErr := decodeByExt(path, try)
        if tryErr == nil {
            return streamer, format, closer, nil
        }
    }
    return nil, beep.Format{}, nil, fmt.Errorf("unsupported audio format %q (supported: %s): %w", ext, supportedFormatsMessage(), err)
}

func decodeByExt(path, ext string) (beep.Streamer, beep.Format, io.Closer, error) {
    switch ext {
    case ".wav", ".wave":
        return decodeBeepFile(path, ext)
    case ".mp3":
        return decodeBeepFile(path, ext)
    case ".flac":
        return decodeBeepFile(path, ext)
    case ".ogg", ".opus":
        return decodeOggFile(path)
    case ".m4a", ".m4b", ".mp4", ".mov", ".m4v", ".aac":
        return decodeAACAsStreamer(path, ext)
    case ".webm":
        return nil, beep.Format{}, nil, fmt.Errorf("webm is not supported yet")
    default:
        return nil, beep.Format{}, nil, fmt.Errorf("unsupported extension %q", ext)
    }
}

func decodeBeepFile(path, ext string) (beep.Streamer, beep.Format, io.Closer, error) {
    f, err := os.Open(path)
    if err != nil {
        return nil, beep.Format{}, nil, err
    }
    var (
        streamer beep.StreamSeekCloser
        format   beep.Format
        decErr   error
    )
    switch ext {
    case ".wav", ".wave":
        streamer, format, decErr = beepwav.Decode(f)
    case ".mp3":
        streamer, format, decErr = mp3.Decode(f)
    case ".flac":
        streamer, format, decErr = flac.Decode(f)
    default:
        f.Close()
        return nil, beep.Format{}, nil, fmt.Errorf("internal: beep decode for %q", ext)
    }
    if decErr != nil {
        f.Close()
        return nil, beep.Format{}, nil, decErr
    }
    return streamer, format, f, nil
}

func decodeAACAsStreamer(path, ext string) (beep.Streamer, beep.Format, io.Closer, error) {
    samples, sr, ch, err := decodeAACPath(path, ext)
    if err != nil {
        return nil, beep.Format{}, nil, err
    }
    if ch <= 0 {
        ch = 1
    }
    return newSamplesStreamer(samples, sr), beep.Format{
        SampleRate:  beep.SampleRate(sr),
        NumChannels: ch,
        Precision:   2,
    }, noopCloser{}, nil
}

type noopCloser struct{}

func (noopCloser) Close() error { return nil }