Feat: 获取&写入 音频文件的 Metadata #43

Merged
um-dev merged 13 commits from feat/audio-meta into master 2022-12-06 17:55:19 +00:00
3 changed files with 263 additions and 11 deletions
Showing only changes of commit 02e065aac4 - Show all commits

View File

@ -4,7 +4,6 @@ import (
"context"
"errors"
"fmt"
"io"
"strings"
"github.com/samber/lo"
@ -102,10 +101,7 @@ func (d *Decoder) GetCoverImage(ctx context.Context) ([]byte, error) {
return nil, fmt.Errorf("qmc[GetCoverImage] extract album art: %w", err)
}
d.cover, err = io.ReadAll(img)
if err != nil {
return nil, fmt.Errorf("qmc[GetCoverImage] read embed cover: %w", err)
}
d.cover = img.Bytes()
return d.cover, nil
}

View File

@ -5,10 +5,14 @@ import (
"context"
"fmt"
"io"
"os"
"os/exec"
"strings"
"unlock-music.dev/cli/algo/common"
)
func ExtractAlbumArt(ctx context.Context, rd io.Reader) (io.Reader, error) {
func ExtractAlbumArt(ctx context.Context, rd io.Reader) (*bytes.Buffer, error) {
cmd := exec.CommandContext(ctx, "ffmpeg",
"-i", "pipe:0", // input from stdin
"-an", // disable audio
@ -21,13 +25,134 @@ func ExtractAlbumArt(ctx context.Context, rd io.Reader) (io.Reader, error) {
stdout, stderr := &bytes.Buffer{}, &bytes.Buffer{}
cmd.Stdout, cmd.Stderr = stdout, stderr
if err := cmd.Start(); err != nil {
if err := cmd.Run(); err != nil {
return nil, fmt.Errorf("ffmpeg run: %w", err)
}
if err := cmd.Wait(); err != nil {
return nil, fmt.Errorf("ffmpeg wait: %w: %s", err, stderr.String())
}
return stdout, nil
}
type UpdateMetadataParams struct {
Audio io.Reader // required
AudioExt string // required
Meta common.AudioMeta // required
AlbumArt io.Reader // optional
AlbumArtExt string // required if AlbumArt is not nil
}
func UpdateAudioMetadata(ctx context.Context, params *UpdateMetadataParams) (*bytes.Buffer, error) {
builder := newFFmpegBuilder()
builder.SetFlag("y") // overwrite output file
out := newOutputBuilder("pipe:1") // use stdout as output
out.AddOption("f", encodeFormatFromExt(params.AudioExt)) // use mp3 muxer
builder.AddOutput(out)
// since ffmpeg doesn't support multiple input streams,
// we need to write the audio to a temp file
audioPath, err := writeTempFile(params.Audio, params.AudioExt)
if err != nil {
return nil, fmt.Errorf("updateAudioMeta write temp file: %w", err)
}
defer os.Remove(audioPath)
// input audio -> output audio
builder.AddInput(newInputBuilder(audioPath)) // input 0: audio
out.AddOption("map", "0:a")
out.AddOption("codec:a", "copy")
// input cover -> output cover
if params.AlbumArt != nil &&
params.AudioExt != ".wav" /* wav doesn't support attached image */ {
// write cover to temp file
artPath, err := writeTempFile(params.AlbumArt, params.AlbumArtExt)
if err != nil {
return nil, fmt.Errorf("updateAudioMeta write temp file: %w", err)
}
defer os.Remove(artPath)
builder.AddInput(newInputBuilder(artPath)) // input 1: cover
out.AddOption("map", "1:v")
switch params.AudioExt {
case ".ogg": // ogg only supports theora codec
out.AddOption("codec:v", "libtheora")
case ".m4a": // .m4a(mp4) requires set codec, disposition, stream metadata
out.AddOption("codec:v", "mjpeg")
out.AddOption("disposition:v", "attached_pic")
out.AddMetadata("s:v", "title", "Album cover")
out.AddMetadata("s:v", "comment", "Cover (front)")
default: // other formats use default behavior
}
}
// set file metadata
album := params.Meta.GetAlbum()
if album != "" {
out.AddMetadata("", "album", album)
}
title := params.Meta.GetTitle()
if album != "" {
out.AddMetadata("", "title", title)
}
artists := params.Meta.GetArtists()
if len(artists) != 0 {
// TODO: it seems that ffmpeg doesn't support multiple artists
out.AddMetadata("", "artist", strings.Join(artists, " / "))
}
// execute ffmpeg
cmd := builder.Command(ctx)
stdout, stderr := &bytes.Buffer{}, &bytes.Buffer{}
cmd.Stdout, cmd.Stderr = stdout, stderr
if err := cmd.Run(); err != nil {
return nil, fmt.Errorf("ffmpeg run: %w", err)
}
return stdout, nil
}
func writeTempFile(rd io.Reader, ext string) (string, error) {
audioFile, err := os.CreateTemp("", "*"+ext)
if err != nil {
return "", fmt.Errorf("ffmpeg create temp file: %w", err)
}
if _, err := io.Copy(audioFile, rd); err != nil {
return "", fmt.Errorf("ffmpeg write temp file: %w", err)
}
if err := audioFile.Close(); err != nil {
return "", fmt.Errorf("ffmpeg close temp file: %w", err)
}
return audioFile.Name(), nil
}
// encodeFormatFromExt returns the file format name the recognized & supporting encoding by ffmpeg.
func encodeFormatFromExt(ext string) string {
switch ext {
case ".flac":
return "flac" // raw FLAC
case ".mp3":
return "mp3" // MP3 (MPEG audio layer 3)
case ".ogg":
return "ogg" // Ogg
case ".m4a":
return "ipod" // iPod H.264 MP4 (MPEG-4 Part 14)
case ".wav":
return "wav" // WAV / WAVE (Waveform Audio)
case ".aac":
return "adts" // ADTS AAC (Advanced Audio Coding)
case ".wma":
return "asf" // ASF (Advanced / Active Streaming Format)
default:
return ""
}
}

131
internal/ffmpeg/options.go Normal file
View File

@ -0,0 +1,131 @@
package ffmpeg
import (
"context"
"os/exec"
"strings"
)
type ffmpegBuilder struct {
binary string // ffmpeg binary path
options map[string]string // global options
inputs []*inputBuilder // input options
outputs []*outputBuilder // output options
}
func newFFmpegBuilder() *ffmpegBuilder {
return &ffmpegBuilder{
binary: "ffmpeg",
options: make(map[string]string),
}
}
func (b *ffmpegBuilder) AddInput(src *inputBuilder) {
b.inputs = append(b.inputs, src)
}
func (b *ffmpegBuilder) AddOutput(dst *outputBuilder) {
b.outputs = append(b.outputs, dst)
}
func (b *ffmpegBuilder) SetBinary(bin string) {
b.binary = bin
}
func (b *ffmpegBuilder) SetFlag(flag string) {
b.options[flag] = ""
}
func (b *ffmpegBuilder) SetOption(name, value string) {
b.options[name] = value
}
func (b *ffmpegBuilder) Args() (args []string) {
for name, val := range b.options {
args = append(args, "-"+name)
if val != "" {
args = append(args, val)
}
}
for _, input := range b.inputs {
args = append(args, input.Args()...)
}
for _, output := range b.outputs {
args = append(args, output.Args()...)
}
return
}
func (b *ffmpegBuilder) Command(ctx context.Context) *exec.Cmd {
bin := "ffmpeg"
if b.binary != "" {
bin = b.binary
}
return exec.CommandContext(ctx, bin, b.Args()...)
}
// inputBuilder is the builder for ffmpeg input options
type inputBuilder struct {
path string
options map[string][]string
}
func newInputBuilder(path string) *inputBuilder {
return &inputBuilder{
path: path,
options: make(map[string][]string),
}
}
func (b *inputBuilder) AddOption(name, value string) {
b.options[name] = append(b.options[name], value)
}
func (b *inputBuilder) Args() (args []string) {
for name, values := range b.options {
for _, val := range values {
args = append(args, "-"+name, val)
}
}
return append(args, "-i", b.path)
}
// outputBuilder is the builder for ffmpeg output options
type outputBuilder struct {
path string
options map[string][]string
}
func newOutputBuilder(path string) *outputBuilder {
return &outputBuilder{
path: path,
options: make(map[string][]string),
}
}
func (b *outputBuilder) AddOption(name, value string) {
b.options[name] = append(b.options[name], value)
}
func (b *outputBuilder) Args() (args []string) {
for name, values := range b.options {
for _, val := range values {
args = append(args, "-"+name, val)
}
}
return append(args, b.path)
}
// AddMetadata is the shortcut for adding "metadata" option
func (b *outputBuilder) AddMetadata(stream, key, value string) {
optVal := strings.TrimSpace(key) + "=" + strings.TrimSpace(value)
if stream != "" {
b.AddOption("metadata:"+stream, optVal)
} else {
b.AddOption("metadata", optVal)
}
}