Merge pull request #115 from kounoike/improvement/transcribe

文字起こし速度の改善
kounoike · Apr 18, 2023 · 4f28658 · 4f28658
2 parents c1651e9 + 102a969
commit 4f28658
Show file tree

Hide file tree

Showing 3 changed files with 3 additions and 3 deletions.
diff --git a/docker-compose/worker/run_whisper.py b/docker-compose/worker/run_whisper.py
@@ -17,7 +17,7 @@
 # or run on CPU with INT8
 # model = WhisperModel(model_size, device="cpu", compute_type="int8")
 
-segments, info = model.transcribe(sys.argv[2], beam_size=5, language="ja", initial_prompt=initial_prompt)
+segments, info = model.transcribe(sys.argv[2], beam_size=5, language="ja", initial_prompt=initial_prompt, vad_filter=True)
 
 #print("Detected language '%s' with probability %f" % (info.language, info.language_probability))
 

diff --git a/tasks/transcription_local_task.go b/tasks/transcription_local_task.go
@@ -77,7 +77,7 @@ func (e *ProgramTranscriberLocal) ProcessTask(ctx context.Context, t *asynq.Task
 	}
 
 	tmpFile := fmt.Sprintf("/tmp/%d.wav", p.ProgramId)
-	commandLine := fmt.Sprintf(`ffmpeg -hide_banner -i "%s" -vn "%s" -y`, inputFile, tmpFile)
+	commandLine := fmt.Sprintf(`ffmpeg -hide_banner -i "%s" -ac 1 -ar 16000 -vn -sn "%s" -y`, inputFile, tmpFile)
 
 	e.logger.Info("Running split audio command", zap.String("command", commandLine))
 

diff --git a/whisper/run_whisper.py b/whisper/run_whisper.py
@@ -18,7 +18,7 @@
 # or run on CPU with INT8
 # model = WhisperModel(model_size, device="cpu", compute_type="int8")
 
-segments, info = model.transcribe(sys.argv[2], beam_size=5, language="ja", initial_prompt=initial_prompt)
+segments, info = model.transcribe(sys.argv[2], beam_size=5, language="ja", initial_prompt=initial_prompt, vad_filter=True)
 
 #print("Detected language '%s' with probability %f" % (info.language, info.language_probability))