lostecho/songyi
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update voice process use gradio

Browse Source
This commit is contained in:
YuanHui
2025-03-05 12:48:47 +08:00
parent 8cf2db5dbc
commit 5b5570ccc8
6 changed files with 86 additions and 57 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
config.ini
View File

@@ -5,4 +5,7 @@ sort=
max_download_threads = 5
max_retry_attempts = 3
authorization_token=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJhdWQiOiIxMDAwMDgzNDciLCJleHAiOjE3NDI2MDc3NDAsImp0aSI6IjEwMjM4YTJmLTBiN2QtNDIwNi1iNDU2LTQ1MTRiMjFjZGM4MyIsImlhdCI6MTc0MDAxNTc0MCwiaXNzIjoiYXBwdXNlciIsInVpZCI6ImJlMmViOGIyLTFhOTItNGVmMC05ZDAwLTA1YTlkN2E2OWRiMiIsInNjaGVtZSI6Imp3dGhzIiwic2lkIjoiZDliZjIzNjQtNjVhYi00ZWNkLThhZjctY2MzMDcxODU0M2M5In0.05GqlG4rhwwlbuQUfEHlHTB-vAz2lOh5JCVlQ6j4V8s
authorization_token=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJhdWQiOiIxMDAwMDgzNDciLCJleHAiOjE3NDI2MDc3NDAsImp0aSI6IjEwMjM4YTJmLTBiN2QtNDIwNi1iNDU2LTQ1MTRiMjFjZGM4MyIsImlhdCI6MTc0MDAxNTc0MCwiaXNzIjoiYXBwdXNlciIsInVpZCI6ImJlMmViOGIyLTFhOTItNGVmMC05ZDAwLTA1YTlkN2E2OWRiMiIsInNjaGVtZSI6Imp3dGhzIiwic2lkIjoiZDliZjIzNjQtNjVhYi00ZWNkLThhZjctY2MzMDcxODU0M2M5In0.05GqlG4rhwwlbuQUfEHlHTB-vAz2lOh5JCVlQ6j4V8s
voice2txt_url=https://api.siliconflow.cn/v1/audio/transcriptions
voice_token=sk-vksrlpckcpttnpjgftpgwytmiipjmvhyzmnffhbhjpahbfiq

26
course.py
View File

@@ -154,8 +154,8 @@ def get_course():
else:
response = requests.get(f'https://bandu-api.songy.info/v2/courses/{course_id}/contents', headers=headers)
contents_data = response.json()
with open(json_filename, 'w', encoding='utf-8') as json_file:
json.dump(contents_data, json_file, ensure_ascii=False, indent=4)
with open(json_filename, 'w', encoding='utf-8') as save_json_file:
json.dump(contents_data, save_json_file, ensure_ascii=False, indent=4)
for item in contents_data['data']:
cursor.execute('''
@@ -242,7 +242,7 @@ def get_course():
folder_path = os.path.join(course_id_folder, folder_name)
if not os.path.exists(folder_path):
os.makedirs(folder_path)
move_file = os.path.join(folder_path, attachment['name']);
move_file = os.path.join(folder_path, attachment['name'])
shutil.move(filename, move_file)
# 保存category为text的content到TXT文件
@@ -253,21 +253,27 @@ def get_course():
txt_file.write(content + '\n')
# 处理mp4文件
mp4_folder = os.path.join(course_id_folder, 'mp4')
mp4_file = None
exist_md_file = False
if os.path.exists(mp4_folder):
# 遍历指定文件夹内的所有文件和子文件夹
for root, dirs, files in os.walk(mp4_folder):
for file in files:
# 检查是否已经存在
if file.lower().endswith('.md'):
exist_md_file = True
# 检查文件扩展名是否为.mp4
if file.lower().endswith('.mp4'):
# 构建完整的 MP4 文件路径
mp4_file = os.path.join(root, file)
# 调用 mp4_to_wav 函数进行转换
wav_file = convert_mp4(mp4_file)
if wav_file is not None:
try:
process_audio_file(wav_file)
except:
print('process_audio_file fail')
if (not exist_md_file) and mp4_file is not None:
# 调用 mp4_to_wav 函数进行转换
wav_file = convert_mp4(mp4_file)
if wav_file is not None:
try:
process_audio_file(wav_file)
except:
print('process_audio_file fail')
if __name__ == '__main__':

BIN
courses.db
View File

Binary file not shown.

2
monitor

Submodule monitor updated: bd25bd3ec0...b1d647b3b8

5
requirements.txt
View File

@@ -1,2 +1,3 @@
requests
pydub
requests~=2.32.3
pydub~=0.25.1
gradio_client~=1.7.2

105
video_voice_process.py
View File

@@ -1,4 +1,5 @@
# -*- coding: utf-8 -*-
import configparser
import logging
import os
import shutil
@@ -6,11 +7,20 @@ import threading
from concurrent.futures import ThreadPoolExecutor, as_completed
import requests
from gradio_client import Client, handle_file
from pydub import AudioSegment
from pydub.silence import split_on_silence
# 创建一个锁对象
file_write_lock = threading.Lock()
use_remote_api = False
process_workers = 5 if use_remote_api else 2
config = configparser.ConfigParser()
config.read('config.ini')
token = config['DEFAULT']['voice_token']
url = config['DEFAULT']['voice2txt_url']
headers = {
"Authorization": f'Bearer {token}'
}
# 配置日志
@@ -44,6 +54,8 @@ def process_audio_file(audio_file_path):
wav_filename = os.path.splitext(os.path.basename(audio_file_path))[0]
# 生成对应的 WAV 文件路径
md_file = os.path.join(wav_dir, f"{wav_filename}.md")
# 创建一个锁对象
file_write_lock = threading.Lock()
try:
audio_file = AudioSegment.from_wav(audio_file_path)
@@ -57,45 +69,11 @@ def process_audio_file(audio_file_path):
else:
new_audio_chunks.append(chunk)
url = "https://api.siliconflow.cn/v1/audio/transcriptions"
headers = {
"Authorization": "Bearer sk-lakndqcjlmtukekcliwkkryaxquifduhvzgcnlhofzvofllv"
}
sentences = [] # 用于存储所有句子的列表
def send_request(chunk, index):
try:
audio_part_path = os.path.join('media', f"{file_name_without_extension}_chunk_{index}.wav")
chunk.export(audio_part_path, format="wav")
logging.info(f'Exported chunk file {audio_part_path} for {file_name_with_extension}')
multipart_form_data = {
'file': (os.path.basename(audio_part_path), open(audio_part_path, 'rb')),
'model': (None, 'FunAudioLLM/SenseVoiceSmall')
}
response = requests.post(url, files=multipart_form_data, headers=headers)
result = response.json()
text = result["text"]
print(text)
return index, text # 返回索引和文本
except Exception as e:
logging.error(f'Error processing {file_name_with_extension}, chunk {index}: {str(e)}')
# 将出错的音频片段复制到error文件夹
error_dir = os.path.join(os.getcwd(), 'media', 'error')
if not os.path.exists(error_dir):
os.makedirs(error_dir)
error_path = os.path.join(error_dir, f"{file_name_without_extension}_chunk_{index}.wav")
shutil.copy(audio_part_path, error_path)
logging.error(f'Copied request failed chunk file {file_name_with_extension} to {error_path}')
return index, "" # 返回空文本
finally:
if os.path.exists(audio_part_path):
os.remove(audio_part_path)
with ThreadPoolExecutor(max_workers=5) as executor:
futures = {executor.submit(send_request, chunk, i): i for i, chunk in enumerate(new_audio_chunks)}
with ThreadPoolExecutor(max_workers=process_workers) as executor:
futures = {executor.submit(send_request, chunk, i, file_name_without_extension): i for i, chunk in
enumerate(new_audio_chunks)}
for future in as_completed(futures):
index = futures[future]
@@ -110,22 +88,63 @@ def process_audio_file(audio_file_path):
for sentence in sorted(sentences, key=lambda x: x[0]): # 根据索引排序
markdown_content += f"{sentence[1]}\n\n"
with file_write_lock: # 确保文件写入操作的线程安全
# with file_write_lock: # 确保文件写入操作的线程安全
# md_file_path = os.path.join('media', file_name_without_extension + '.md')
with open(md_file, "w", encoding="utf-8") as f:
f.write(markdown_content)
with open(md_file, "w", encoding="utf-8") as f:
f.write(markdown_content)
logging.info(f"Finished processing {file_name_with_extension}")
except Exception as e:
logging.error(f"Failed to process {file_name_with_extension}: {str(e)}")
def send_request(chunk, index, file_name_without_extension):
audio_part_path = os.path.join('media', f"{file_name_without_extension}_chunk_{index}.wav")
chunk.export(audio_part_path, format="wav")
logging.info(f'Exported chunk file {audio_part_path} for {file_name_without_extension}')
try:
if use_remote_api:
multipart_form_data = {
'file': (os.path.basename(audio_part_path), open(audio_part_path, 'rb')),
'model': (None, 'FunAudioLLM/SenseVoiceSmall')
}
response = requests.post(url, files=multipart_form_data, headers=headers)
result = response.json()
text = result["text"]
print(text)
else:
client = Client("http://192.168.31.3:7860/")
text = client.predict(
input_wav=handle_file(audio_part_path),
language="zh",
api_name="/model_inference"
)
print(text)
return index, text # 返回索引和文本
except Exception as e:
logging.error(f'Error processing {file_name_without_extension}, chunk {index}: {str(e)}')
# 将出错的音频片段复制到error文件夹
error_dir = os.path.join(os.getcwd(), 'media', 'error')
if not os.path.exists(error_dir):
os.makedirs(error_dir)
error_path = os.path.join(error_dir, f"{file_name_without_extension}_chunk_{index}.wav")
shutil.copy(audio_part_path, error_path)
logging.error(f'Copied request failed chunk file {file_name_without_extension} to {error_path}')
return index, "" # 返回空文本
finally:
if os.path.exists(audio_part_path):
os.remove(audio_part_path)
def main():
all_files = os.listdir('media')
audio_files = [file for file in all_files if file.endswith('.wav')]
print(audio_files)
with ThreadPoolExecutor(max_workers=5) as executor:
with ThreadPoolExecutor(max_workers=process_workers) as executor:
for audio_file in audio_files:
audio_file_path = os.path.join('media', audio_file)
executor.submit(process_audio_file, audio_file_path)