go-whisper-api/transcode/ogg_decode.go

package transcode

import (
    "bytes"
    "errors"
    "fmt"
    "io"
    "os"
    "strings"

    "github.com/gopxl/beep"
    "github.com/gopxl/beep/vorbis"
    "github.com/pion/opus"
    "github.com/pion/opus/pkg/oggreader"
)

func decodeOggFile(path string) (beep.Streamer, beep.Format, io.Closer, error) {
    switch sniffOggCodec(path) {
    case "opus":
        return decodeOggOpus(path)
    case "vorbis":
        return decodeOggVorbis(path)
    default:
        streamer, format, closer, err := decodeOggVorbis(path)
        if err == nil {
            return streamer, format, closer, nil
        }
        if isVorbisInvalidHeader(err) {
            return decodeOggOpus(path)
        }
        return nil, beep.Format{}, nil, err
    }
}

func sniffOggCodec(path string) string {
    f, err := os.Open(path)
    if err != nil {
        return ""
    }
    defer f.Close()
    buf := make([]byte, 8192)
    n, _ := io.ReadFull(f, buf)
    buf = buf[:n]
    if len(buf) < 4 || !bytes.HasPrefix(buf, []byte("OggS")) {
        return ""
    }
    if bytes.Contains(buf, []byte("OpusHead")) {
        return "opus"
    }
    // Vorbis ID packet: 0x01 + "vorbis"
    if bytes.Contains(buf, []byte{0x01, 'v', 'o', 'r', 'b', 'i', 's'}) {
        return "vorbis"
    }
    return ""
}

func isVorbisInvalidHeader(err error) bool {
    if err == nil {
        return false
    }
    msg := strings.ToLower(err.Error())
    return strings.Contains(msg, "invalid header") || strings.Contains(msg, "vorbis:")
}

func decodeOggVorbis(path string) (beep.Streamer, beep.Format, io.Closer, error) {
    f, err := os.Open(path)
    if err != nil {
        return nil, beep.Format{}, nil, err
    }
    streamer, format, decErr := vorbis.Decode(f)
    if decErr != nil {
        f.Close()
        return nil, beep.Format{}, nil, fmt.Errorf("ogg/vorbis: %w", decErr)
    }
    return streamer, format, f, nil
}

func decodeOggOpus(path string) (beep.Streamer, beep.Format, io.Closer, error) {
    f, err := os.Open(path)
    if err != nil {
        return nil, beep.Format{}, nil, err
    }
    ogg, header, err := oggreader.NewWith(f)
    if err != nil {
        f.Close()
        return nil, beep.Format{}, nil, fmt.Errorf("ogg/opus: %w", err)
    }

    sr := int(header.SampleRate)
    if sr <= 0 {
        sr = 48000
    }
    ch := int(header.Channels)
    if ch <= 0 {
        ch = 1
    }

    dec, err := opus.NewDecoderWithOutput(sr, ch)
    if err != nil {
        f.Close()
        return nil, beep.Format{}, nil, fmt.Errorf("ogg/opus decoder: %w", err)
    }

    const maxFrameSamples = 5760
    pcmBuf := make([]float32, maxFrameSamples*ch)
    var samples []float64

    for {
        pkt, _, err := ogg.ParseNextPacket()
        if errors.Is(err, io.EOF) {
            break
        }
        if err != nil {
            f.Close()
            return nil, beep.Format{}, nil, fmt.Errorf("ogg/opus read: %w", err)
        }
        if len(pkt) == 0 || bytes.HasPrefix(pkt, []byte("OpusHead")) || bytes.HasPrefix(pkt, []byte("OpusTags")) {
            continue
        }

        n, err := dec.DecodeToFloat32(pkt, pcmBuf)
        if err != nil {
            f.Close()
            return nil, beep.Format{}, nil, fmt.Errorf("ogg/opus decode: %w", err)
        }
        if n <= 0 {
            continue
        }
        total := n * ch
        if total > len(pcmBuf) {
            total = len(pcmBuf)
        }
        for i := 0; i < total; i++ {
            samples = append(samples, float64(pcmBuf[i]))
        }
    }

    if len(samples) == 0 {
        f.Close()
        return nil, beep.Format{}, nil, fmt.Errorf("ogg/opus: no audio samples")
    }

    outCh := ch
    if outCh > 1 {
        samples = interleavedToMono(samples, outCh)
        outCh = 1
    }

    return newSamplesStreamer(samples, sr), beep.Format{
        SampleRate:  beep.SampleRate(sr),
        NumChannels: outCh,
        Precision:   2,
    }, f, nil
}