Merge pull request #37 from Eisaichen/main

Workaround for crash on Windows
Chenyme · Jul 1, 2024 · 2436ecc · 2436ecc
2 parents af0b103 + f04ed6c
commit 2436ecc
Show file tree

Hide file tree

Showing 3 changed files with 19 additions and 8 deletions.
diff --git a/project/content.py b/project/content.py
@@ -6,7 +6,7 @@
 import streamlit as st
 import streamlit_antd_components as sac
 from openai import OpenAI
-from .utils.utils2 import (openai_whisper_result, faster_whisper_result, file_to_mp3, check_ffmpeg, check_cuda_support)
+from .utils.utils2 import (openai_whisper_result, runWhisperSeperateProc, file_to_mp3, check_ffmpeg, check_cuda_support)
 
 
 def content():
@@ -203,7 +203,7 @@ def content():
                         model = faster_whisper_model
                         if faster_whisper_local:
                             model = faster_whisper_local_path
-                        result = faster_whisper_result(output_file, device, model, "Don’t make each line too long.", temperature_setting, False, "自动识别", beam_size_setting, 500)
+                        result = runWhisperSeperateProc(output_file, device, model, "Don’t make each line too long.", temperature_setting, False, "自动识别", beam_size_setting, 500)
                     st.session_state.text = result["text"]
                     st.toast("已识别完成，开始对话叭！", icon=":material/task_alt:")
                 else:

diff --git a/project/media.py b/project/media.py
@@ -6,7 +6,7 @@
 import datetime
 import streamlit as st
 import streamlit_antd_components as sac
-from project.utils.utils2 import (file_to_mp3, openai_whisper_result, faster_whisper_result, translate, local_translate,
+from project.utils.utils2 import (file_to_mp3, openai_whisper_result, runWhisperSeperateProc, translate, local_translate,
                                   generate_srt_from_result, generate_srt_from_result_2, srt_mv, show_video, parse_srt_file,
                                   convert_to_srt, srt_to_ass, srt_to_stl, srt_to_vtt, check_cuda_support, check_ffmpeg,
                                   add_font_settings)
@@ -391,7 +391,7 @@ def media():
                             model = faster_whisper_model
                             if faster_whisper_local:
                                 model = faster_whisper_local_path
-                            result = faster_whisper_result(output_file1, device, model, whisper_prompt_setting, temperature_setting, vad_setting, lang_setting, beam_size_setting, min_vad_setting)
+                            result = runWhisperSeperateProc(output_file1, device, model, whisper_prompt_setting, temperature_setting, vad_setting, lang_setting, beam_size_setting, min_vad_setting)
 
                         time3 = time.time()
                         translation_dict = {
@@ -526,7 +526,7 @@ def media():
                             model = faster_whisper_model
                             if faster_whisper_local:
                                 model = faster_whisper_local_path
-                            result = faster_whisper_result(output_file, device, model, whisper_prompt_setting, temperature_setting, vad_setting, lang_setting, beam_size_setting, min_vad_setting)
+                            result = runWhisperSeperateProc(output_file, device, model, whisper_prompt_setting, temperature_setting, vad_setting, lang_setting, beam_size_setting, min_vad_setting)
 
                         time3 = time.time()
                         translation_dict = {
@@ -635,7 +635,7 @@ def media():
                             model = faster_whisper_model
                             if faster_whisper_local:
                                 model = faster_whisper_local_path
-                            result = faster_whisper_result(output_file, device, model, whisper_prompt_setting, temperature_setting, vad_setting, lang_setting, beam_size_setting, min_vad_setting)
+                            result = runWhisperSeperateProc(output_file, device, model, whisper_prompt_setting, temperature_setting, vad_setting, lang_setting, beam_size_setting, min_vad_setting)
 
                         time3 = time.time()
                         translation_dict = {

diff --git a/project/utils/utils2.py b/project/utils/utils2.py
@@ -1,3 +1,4 @@
+from multiprocessing import Process, Manager
 import os
 import re
 import math
@@ -114,7 +115,7 @@ def openai_whisper_result(key, base, path, prompt, temperature):
     return result
 
 
-def faster_whisper_result(file_path, device, model_name, prompt, temp, vad, lang, beam_size, min_vad):
+def faster_whisper_result(file_path, device, model_name, prompt, temp, vad, lang, beam_size, min_vad, returnList):
     if model_name not in ['tiny', 'tiny.en', 'base', 'base.en', 'small', 'small.en', 'medium', 'medium.en', 'large-v1',
                           'large-v2', 'large-v3', 'large', 'distil-small.en', 'distil-medium.en', 'distil-large-v2',
                           'distil-large-v3']:
@@ -165,7 +166,17 @@ def faster_whisper_result(file_path, device, model_name, prompt, temp, vad, lang
 
     result = faster_whisper_result_dict(segments)
     print(f"- whisper识别内容：\n{result['text']}\n")
-    return result
+    returnList[0] = result
+
+
+def runWhisperSeperateProc(*args):
+    with Manager() as manager:
+        returnList = manager.list([None])
+        p = Process(target=faster_whisper_result, args=args + (returnList,))
+        p.start()
+        p.join()
+        p.close()
+        return returnList[0]
 
 
 def translate(system_prompt, user_prompt, api_key, base_url, model, result, wait_time, srt):