red-future/media
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

代码初始化

Browse Source
This commit is contained in:
李明康
2026-05-20 11:32:39 +08:00
parent 219b7e39c7
commit e76bf57d54
20 changed files with 1585 additions and 309 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
service/asr/whisper_service.go
View File

@@ -9,6 +9,7 @@ import (
"os"
"os/exec"
"path/filepath"
"runtime"
"strings"
"time"
@@ -24,11 +25,10 @@ const (
backendCpp // whisper.cpp (whisper-cpp)
)
// WhisperService 语音识别服务
type WhisperService struct{}
type whisperService struct{}
// Whisper 语音识别服务单例
var Whisper = new(WhisperService)
var Whisper = new(whisperService)
// TranscribeReq 语音识别请求
type TranscribeReq struct {
@@ -54,7 +54,7 @@ type Segment struct {
}
// Transcribe 对音频文件进行语音识别（自动检测后端，自动降级）
func (s *WhisperService) Transcribe(ctx context.Context, req *TranscribeReq) (res *TranscribeRes, err error) {
func (s *whisperService) Transcribe(ctx context.Context, req *TranscribeReq) (res *TranscribeRes, err error) {
// 1. 校验音频文件
if _, err = os.Stat(req.AudioPath); os.IsNotExist(err) {
return nil, fmt.Errorf("音频文件不存在: %s", req.AudioPath)
@@ -94,7 +94,7 @@ func (s *WhisperService) Transcribe(ctx context.Context, req *TranscribeReq) (re
}
// transcribeWithCLI 使用 whisper CLI 命令
func (s *WhisperService) transcribeWithCLI(ctx context.Context, req *TranscribeReq, whisperPath, model, language string) (res *TranscribeRes, err error) {
func (s *whisperService) transcribeWithCLI(ctx context.Context, req *TranscribeReq, whisperPath, model, language string) (res *TranscribeRes, err error) {
outputDir := filepath.Dir(req.AudioPath)
modelDir := g.Cfg().MustGet(ctx, "whisper.model_dir", "").String()
threads := g.Cfg().MustGet(ctx, "whisper.threads", 2).Int()
@@ -122,7 +122,7 @@ func (s *WhisperService) transcribeWithCLI(ctx context.Context, req *TranscribeR
}
// transcribeWithPython 使用 python -m whisper
func (s *WhisperService) transcribeWithPython(ctx context.Context, req *TranscribeReq, model, language string) (res *TranscribeRes, err error) {
func (s *whisperService) transcribeWithPython(ctx context.Context, req *TranscribeReq, model, language string) (res *TranscribeRes, err error) {
// 查找 python
pythonPath, err := exec.LookPath("python3")
if err != nil {
@@ -160,7 +160,7 @@ func (s *WhisperService) transcribeWithPython(ctx context.Context, req *Transcri
}
// readTxtResult 读取 whisper 输出的 txt 文件
func (s *WhisperService) readTxtResult(outputDir, audioPath, model string) (res *TranscribeRes, err error) {
func (s *whisperService) readTxtResult(outputDir, audioPath, model string) (res *TranscribeRes, err error) {
baseName := strings.TrimSuffix(filepath.Base(audioPath), filepath.Ext(audioPath))
txtPaths := []string{
filepath.Join(outputDir, baseName+".txt"),
@@ -201,7 +201,7 @@ func cleanTranscript(text string) string {
}
// detectBackend 检测可用的 whisper 后端，返回后端类型和可执行路径
func (s *WhisperService) detectBackend() (WhisperBackend, string) {
func (s *whisperService) detectBackend() (WhisperBackend, string) {
// 1. 优先检测 C++ 版 whisper.cpp（最快，但参数格式不同）
for _, name := range []string{"whisper-cpp", "whisper-cli"} {
if path, err := exec.LookPath(name); err == nil {
@@ -242,7 +242,7 @@ func (s *WhisperService) detectBackend() (WhisperBackend, string) {
}
// resolveCppModelPath 查找或下载 whisper.cpp 模型文件
func (s *WhisperService) resolveCppModelPath(model string) string {
func (s *whisperService) resolveCppModelPath(model string) string {
modelName := strings.TrimPrefix(model, "ggml-")
modelName = strings.TrimSuffix(modelName, ".bin")
@@ -262,8 +262,20 @@ func (s *WhisperService) resolveCppModelPath(model string) string {
altPaths := []string{
cppModelName,
filepath.Join(home, ".cache", "whisper", "ggml-"+modelName+"-q5_0.bin"),
"/opt/homebrew/share/whisper-cpp/models/" + cppModelName,
"/usr/local/share/whisper-cpp/models/" + cppModelName,
}
// macOS: Homebrew 安装的 whisper.cpp 模型路径
if runtime.GOOS == "darwin" {
altPaths = append(altPaths,
"/opt/homebrew/share/whisper-cpp/models/"+cppModelName,
"/usr/local/share/whisper-cpp/models/"+cppModelName,
)
}
// Linux: 常见系统安装路径
if runtime.GOOS == "linux" {
altPaths = append(altPaths,
"/usr/share/whisper-cpp/models/"+cppModelName,
"/usr/local/share/whisper-cpp/models/"+cppModelName,
)
}
for _, p := range altPaths {
if _, err := os.Stat(p); err == nil {
@@ -310,7 +322,7 @@ func (s *WhisperService) resolveCppModelPath(model string) string {
}
// downloadFile 下载文件到指定路径（支持超时）
func (s *WhisperService) downloadFile(url, destPath string, timeout time.Duration) error {
func (s *whisperService) downloadFile(url, destPath string, timeout time.Duration) error {
tmpPath := destPath + ".tmp"
out, err := os.Create(tmpPath)
if err != nil {
@@ -346,7 +358,7 @@ func (s *WhisperService) downloadFile(url, destPath string, timeout time.Duratio
}
// transcribeWithCpp 使用 whisper.cpp（C++ 版，参数格式不同）
func (s *WhisperService) transcribeWithCpp(ctx context.Context, req *TranscribeReq, binaryPath, model, language string) (res *TranscribeRes, err error) {
func (s *whisperService) transcribeWithCpp(ctx context.Context, req *TranscribeReq, binaryPath, model, language string) (res *TranscribeRes, err error) {
outputDir := filepath.Dir(req.AudioPath)
baseName := strings.TrimSuffix(filepath.Base(req.AudioPath), filepath.Ext(req.AudioPath))
outputPrefix := filepath.Join(outputDir, baseName)