Skip to content

Formerly KrillinAI, VideoTranslation and Dubbing Tool for 120+ Languages. Open-source AI workspace for creators, powered by Codex. Video downloader, video translation, create videos, images, audios, and edits with Agents in one place. For your Social Media like YouTube, TikTok, Instagram, etc.

feat: add gpu acceleration option for fasterwhisper

puji4810a11a4240a397
diff --git a/config/config.go b/config/config.go
index 244f82f0..d643238e 100644
--- a/config/config.go
+++ b/config/config.go
@@ -59,12 +59,13 @@ type AliyunTranscribeConfig struct {
}
type Transcribe struct {
- Provider string `toml:"provider"`
- Openai OpenaiCompatibleConfig `toml:"openai"`
- Fasterwhisper LocalModelConfig `toml:"fasterwhisper"`
- Whisperkit LocalModelConfig `toml:"whisperkit"`
- Whispercpp LocalModelConfig `toml:"whispercpp"`
- Aliyun AliyunTranscribeConfig `toml:"aliyun"`
+ Provider string `toml:"provider"`
+ EnableGpuAcceleration bool `toml:"enable_gpu_acceleration"`
+ Openai OpenaiCompatibleConfig `toml:"openai"`
+ Fasterwhisper LocalModelConfig `toml:"fasterwhisper"`
+ Whisperkit LocalModelConfig `toml:"whisperkit"`
+ Whispercpp LocalModelConfig `toml:"whispercpp"`
+ Aliyun AliyunTranscribeConfig `toml:"aliyun"`
}
type AliyunTtsConfig struct {
@@ -108,7 +109,8 @@ var Conf = Config{
Model: "gpt-4o-mini",
},
Transcribe: Transcribe{
- Provider: "openai",
+ Provider: "openai",
+ EnableGpuAcceleration: false, // 默认不开启GPU加速
Openai: OpenaiCompatibleConfig{
Model: "whisper-1",
},
diff --git a/internal/desktop/ui.go b/internal/desktop/ui.go
index 17682d88..ed65a61b 100644
--- a/internal/desktop/ui.go
+++ b/internal/desktop/ui.go
@@ -315,6 +315,7 @@ func createTranscribeConfigGroup() *fyne.Container {
form := widget.NewForm(
widget.NewFormItem("提供商 Provider", providerSelect),
+ widget.NewFormItem("GPU加速 GPU acceleration", widget.NewCheckWithData("启用 Enable", binding.BindBool(&config.Conf.Transcribe.EnableGpuAcceleration))),
widget.NewFormItem("OpenAI Base URL", openaiBaseUrlEntry),
widget.NewFormItem("OpenAI API Key", openaiApiKeyEntry),
diff --git a/internal/handler/config.go b/internal/handler/config.go
index eb2ae3c1..4201dd72 100644
--- a/internal/handler/config.go
+++ b/internal/handler/config.go
@@ -33,8 +33,9 @@ type ConfigRequest struct {
Model string `json:"model"`
} `json:"llm"`
Transcribe struct {
- Provider string `json:"provider"`
- Openai struct {
+ Provider string `json:"provider"`
+ EnableGpuAcceleration bool `json:"enableGpuAcceleration"`
+ Openai struct {
BaseUrl string `json:"baseUrl"`
ApiKey string `json:"apiKey"`
Model string `json:"model"`
@@ -126,6 +127,7 @@ func (h Handler) GetConfig(c *gin.Context) {
// 转录配置
configResponse.Transcribe.Provider = config.Conf.Transcribe.Provider
+ configResponse.Transcribe.EnableGpuAcceleration = config.Conf.Transcribe.EnableGpuAcceleration
configResponse.Transcribe.Openai.BaseUrl = config.Conf.Transcribe.Openai.BaseUrl
configResponse.Transcribe.Openai.ApiKey = config.Conf.Transcribe.Openai.ApiKey
configResponse.Transcribe.Openai.Model = config.Conf.Transcribe.Openai.Model
@@ -199,6 +201,7 @@ func (h Handler) UpdateConfig(c *gin.Context) {
// 更新转录配置
config.Conf.Transcribe.Provider = req.Transcribe.Provider
+ config.Conf.Transcribe.EnableGpuAcceleration = req.Transcribe.EnableGpuAcceleration
config.Conf.Transcribe.Openai.BaseUrl = req.Transcribe.Openai.BaseUrl
config.Conf.Transcribe.Openai.ApiKey = req.Transcribe.Openai.ApiKey
config.Conf.Transcribe.Openai.Model = req.Transcribe.Openai.Model
diff --git a/pkg/fasterwhisper/transcription.go b/pkg/fasterwhisper/transcription.go
index d5e19827..e5565069 100644
--- a/pkg/fasterwhisper/transcription.go
+++ b/pkg/fasterwhisper/transcription.go
@@ -2,7 +2,7 @@ package fasterwhisper
import (
"encoding/json"
- "go.uber.org/zap"
+ "krillin-ai/config"
"krillin-ai/internal/storage"
"krillin-ai/internal/types"
"krillin-ai/log"
@@ -10,6 +10,8 @@ import (
"os"
"os/exec"
"strings"
+
+ "go.uber.org/zap"
)
func (c *FastwhisperProcessor) Transcription(audioFile, language, workDir string) (*types.TranscriptionData, error) {
@@ -22,6 +24,12 @@ func (c *FastwhisperProcessor) Transcription(audioFile, language, workDir string
"--output_dir", workDir,
audioFile,
}
+
+ if config.Conf.Transcribe.EnableGpuAcceleration {
+ cmdArgs = append(cmdArgs[:len(cmdArgs)-1], "--compute_type", "float16", cmdArgs[len(cmdArgs)-1])
+ log.GetLogger().Info("FastwhisperProcessor启用GPU加速", zap.String("model", c.Model))
+ }
+
cmd := exec.Command(storage.FasterwhisperPath, cmdArgs...)
log.GetLogger().Info("FastwhisperProcessor转录开始", zap.String("cmd", cmd.String()))
output, err := cmd.CombinedOutput()
diff --git a/static/index.html b/static/index.html
index 1a49c515..7a0628a6 100644
--- a/static/index.html
+++ b/static/index.html
@@ -803,6 +803,16 @@
<option value="aliyun">阿里云</option>
</select>
</div>
+ <div class="form-group">
+ <label class="form-label">
+ GPU加速 Enable Gpu Acceleration
+ <span class="hint">仅 FasterWhisper 生效(Only FasterWhisper)</span>
+ </label>
+ <select id="gpu-acceleration" class="form-select">
+ <option value="true">是 Yes</option>
+ <option value="false">否 No</option>
+ </select>
+ </div>
<div class="form-group">
<label class="form-label">OpenAI Base URL:</label>
<input
@@ -1158,6 +1168,10 @@
// 转录配置
if (configData.transcribe) {
+ document.getElementById("gpu-acceleration").value =
+ configData.transcribe.enableGpuAcceleration === true
+ ? "true"
+ : "false";
document.getElementById("transcribe-provider").value =
configData.transcribe.provider || "openai";
if (configData.transcribe.openai) {
@@ -1270,6 +1284,8 @@
},
transcribe: {
provider: document.getElementById("transcribe-provider").value,
+ enableGpuAcceleration:
+ document.getElementById("gpu-acceleration").value == true,
openai: {
baseUrl: document.getElementById("openai-base-url").value,
apiKey: document.getElementById("openai-api-key").value,