This commit is contained in:
Ubuntu
2026-07-03 16:58:28 +08:00
commit 32b43f85bd
6240 changed files with 1987678 additions and 0 deletions
+40
View File
@@ -0,0 +1,40 @@
# Python
__pycache__/
*.py[cod]
*$py.class
*.so
.Python
*.egg-info/
dist/
build/
# Virtual environments
.venv/
.venv-*/
venv/
env/
# Node dependencies
node_modules/
# 上传的音频文件
uploads/*.webm
uploads/*.wav
uploads/*.mp3
!uploads/.gitkeep
# SSL 证书
cert.pem
key.pem
# 模型文件
models/*.zip
models/*.tar.gz
models/cn-model/
# 日志
*.log
# 临时文件
*.tmp
.DS_Store
+123
View File
@@ -0,0 +1,123 @@
# 🎙️ 音频录音工具
前后端分离的录音应用,支持手机浏览器录音、回放和上传。
## 📁 项目结构
```
audio-recorder/
├── frontend/ # 前端(纯HTML/JS)
│ └── index.html # 录音页面
├── backend/ # 后端(Flask API)
│ ├── app.py # 主服务
│ └── requirements.txt
├── uploads/ # 音频文件存储目录
└── start.sh # 启动脚本
```
## ✨ 功能特性
- ✅ **开始录音** - 点击按钮开始录制音频
- ✅ **结束录音** - 停止录制并保存
- ✅ **回放录音** - 本地播放刚录制的音频
- ✅ **自动上传** - 录音完成后自动上传到服务器
- ✅ **录音计时** - 实时显示录音时长
- ✅ **手机适配** - 响应式设计,支持移动端
## 🚀 快速启动
### 1. 启动服务
```bash
cd /home/ubuntu/nanocode/audio-recorder
chmod +x start.sh
./start.sh
```
### 2. 访问应用
- 本机访问: http://localhost:8890
- 局域网访问: http://<服务器IP>:8890
## 📱 手机浏览器要求
**重要**:手机浏览器录音需要 **HTTPS** 环境
### 解决方案:
#### 方案1:使用 nginx + SSL(推荐)
```nginx
server {
listen 443 ssl;
server_name your-domain.com;
ssl_certificate /path/to/cert.pem;
ssl_certificate_key /path/to/key.pem;
location / {
proxy_pass http://127.0.0.1:8890;
proxy_set_header Host $host;
}
}
```
#### 方案2:本地测试(Chrome)
```bash
# 启动带SSL的隧道(需要安装 localtunnel 或 ngrok)
npx localtunnel --port 8890
```
#### 方案3:内网穿透
使用 ngrok:
```bash
ngrok http 8890
```
## 🔌 API 接口
| 接口 | 方法 | 说明 |
|------|------|------|
| `/api/upload` | POST | 上传音频文件 |
| `/api/recordings` | GET | 列出所有录音 |
| `/api/recordings/<filename>` | GET | 获取单个录音 |
| `/api/recordings/<filename>` | DELETE | 删除录音 |
| `/api/health` | GET | 健康检查 |
## 🛠️ 技术栈
**前端:**
- 原生 HTML5 + CSS3 + JavaScript
- MediaRecorder API(录音)
- Web Audio API(回放)
**后端:**
- Python 3.8+
- Flask
- Flask-CORS
**音频格式:**
- 录制:WebM/Opus(默认)或 MP4
- 存储:WebM 格式
- 兼容性:Chrome, Firefox, Safari, Edge
## ⚠️ 注意事项
1. **权限**:首次使用需要授权麦克风权限
2. **HTTPS**:手机浏览器必须在 HTTPS 环境下才能录音
3. **格式**:不同浏览器支持的音频格式不同(优先 WebM)
4. **存储**:录音文件保存在 `uploads/` 目录
## 📝 后续开发计划
- [ ] 语音转文字(ASR)
- [ ] 录音列表管理
- [ ] 大模型总结
- [ ] 用户认证
- [ ] 录音剪辑
---
创建时间:2026-03-29
+589
View File
@@ -0,0 +1,589 @@
#!/usr/bin/env python3
"""
录音后端服务 - 语音转文字 (FunASR + Vosk 双引擎)
优先 FunASR(阿里达摩院,中文识别效果好),失败回退 Vosk(轻量离线)
"""
from flask import Flask, request, jsonify, send_from_directory
from flask_cors import CORS
from werkzeug.utils import secure_filename
import os
import ssl
import subprocess
import wave
import json
import warnings
from datetime import datetime
warnings.filterwarnings('ignore')
app = Flask(__name__)
CORS(app)
# 配置
UPLOAD_FOLDER = '/home/ubuntu/nanocode/audio-recorder/uploads'
VOSK_MODEL_PATH = '/home/ubuntu/nanocode/audio-recorder/models/cn-model'
ALLOWED_EXTENSIONS = {'webm', 'mp4', 'wav', 'mp3', 'ogg'}
MAX_CONTENT_LENGTH = 100 * 1024 * 1024 # 100MB
app.config['MAX_CONTENT_LENGTH'] = MAX_CONTENT_LENGTH
os.makedirs(UPLOAD_FOLDER, exist_ok=True)
# 全局模型(延迟加载)
funasr_model = None
vosk_model = None
def get_funasr_model():
"""加载 FunASR 模型(阿里,中文识别效果好)"""
global funasr_model
if funasr_model is None:
try:
from funasr import AutoModel
print("🔄 加载 FunASR 模型(阿里达摩院)...")
funasr_model = AutoModel(
model="paraformer-zh",
vad_model="fsmn-vad",
punc_model="ct-punc",
)
print("✅ FunASR 模型加载完成")
except Exception as e:
print(f"⚠️ FunASR 加载失败: {e}")
return None
return funasr_model
def get_vosk_model():
"""加载 Vosk 模型(轻量离线,备用)"""
global vosk_model
if vosk_model is None:
try:
from vosk import Model
if os.path.exists(VOSK_MODEL_PATH):
print(f"🔄 加载 Vosk 模型...")
vosk_model = Model(VOSK_MODEL_PATH)
print("✅ Vosk 模型加载完成")
else:
print(f"⚠️ Vosk 模型不存在: {VOSK_MODEL_PATH}")
return None
except Exception as e:
print(f"⚠️ Vosk 加载失败: {e}")
return None
return vosk_model
def allowed_file(filename):
return '.' in filename and filename.rsplit('.', 1)[1].lower() in ALLOWED_EXTENSIONS
def convert_to_wav(input_path, output_path):
"""将音频转换为 WAV 格式(16kHz, 单声道)"""
try:
cmd = [
'ffmpeg', '-y', '-i', input_path,
'-ar', '16000', '-ac', '1', '-f', 'wav',
output_path
]
subprocess.run(cmd, check=True, capture_output=True, timeout=30)
return os.path.exists(output_path)
except Exception as e:
print(f"转换失败: {e}")
return False
def transcribe_with_funasr(wav_path):
"""使用 FunASR 转写(首选)"""
model = get_funasr_model()
if not model:
return None, "FunASR 模型未加载"
try:
res = model.generate(input=wav_path, batch_size_s=300)
text = res[0]["text"] if res and len(res) > 0 else ""
return text, None
except Exception as e:
return None, f"FunASR 错误: {str(e)}"
def transcribe_with_vosk(wav_path):
"""使用 Vosk 转写(备用)"""
from vosk import KaldiRecognizer
model = get_vosk_model()
if not model:
return None, "Vosk 模型未加载"
try:
wf = wave.open(wav_path, 'rb')
if wf.getnchannels() != 1 or wf.getsampwidth() != 2 or wf.getcomptype() != 'NONE':
return None, "音频格式不支持"
rec = KaldiRecognizer(model, wf.getframerate())
rec.SetWords(True)
results = []
while True:
data = wf.readframes(4000)
if len(data) == 0:
break
if rec.AcceptWaveform(data):
result = json.loads(rec.Result())
if result.get('text'):
results.append(result['text'])
final_result = json.loads(rec.FinalResult())
if final_result.get('text'):
results.append(final_result['text'])
wf.close()
return ' '.join(results), None
except Exception as e:
return None, f"Vosk 错误: {str(e)}"
def transcribe_audio_file(audio_path):
"""
转写音频文件
策略:先尝试 FunASR(效果好),失败则回退 Vosk
"""
# 转换为 WAV
wav_path = audio_path + '.wav'
if not convert_to_wav(audio_path, wav_path):
return None, "音频格式转换失败"
try:
# 1. 优先尝试 FunASR
text, error = transcribe_with_funasr(wav_path)
if text:
return text, "funasr"
# 2. 回退到 Vosk
print(f"⚠️ FunASR 失败({error}),回退到 Vosk...")
text, error = transcribe_with_vosk(wav_path)
if text:
return text, "vosk"
return None, error or "识别失败"
finally:
# 清理临时文件
if os.path.exists(wav_path):
os.remove(wav_path)
@app.route('/api/upload', methods=['POST'])
def upload_audio():
"""接收音频文件上传"""
if 'audio' not in request.files:
return jsonify({'error': '没有文件'}), 400
file = request.files['audio']
if file.filename == '':
return jsonify({'error': '文件名为空'}), 400
if file and allowed_file(file.filename):
timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
filename = f"recording_{timestamp}.webm"
filepath = os.path.join(UPLOAD_FOLDER, filename)
file.save(filepath)
file_size = os.path.getsize(filepath)
return jsonify({
'success': True,
'filename': filename,
'filepath': filepath,
'size': file_size,
'size_human': f"{file_size / 1024:.1f} KB",
'timestamp': timestamp
}), 200
return jsonify({'error': '不支持的文件类型'}), 400
@app.route('/api/transcribe', methods=['POST'])
def transcribe():
"""语音转文字 API"""
if 'audio' not in request.files:
return jsonify({'success': False, 'error': '没有音频文件'}), 400
file = request.files['audio']
if file.filename == '':
return jsonify({'success': False, 'error': '文件名为空'}), 400
try:
# 保存上传的音频
timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
filename = f"recording_{timestamp}.webm"
filepath = os.path.join(UPLOAD_FOLDER, filename)
file.save(filepath)
# 转写
text, engine = transcribe_audio_file(filepath)
if text is None:
return jsonify({
'success': False,
'error': engine, # 错误信息
'filename': filename
}), 500
return jsonify({
'success': True,
'text': text or '(未识别到语音)',
'engine': engine, # 显示使用的引擎
'filename': filename
})
except Exception as e:
print(f"转写异常: {e}")
return jsonify({
'success': False,
'error': str(e)
}), 500
@app.route('/api/recordings', methods=['GET'])
def list_recordings():
"""列出所有录音文件"""
recordings = []
for filename in sorted(os.listdir(UPLOAD_FOLDER), reverse=True):
if allowed_file(filename):
filepath = os.path.join(UPLOAD_FOLDER, filename)
stat = os.stat(filepath)
recordings.append({
'filename': filename,
'size': stat.st_size,
'size_human': f"{stat.st_size / 1024:.1f} KB",
'created': datetime.fromtimestamp(stat.st_mtime).isoformat()
})
return jsonify({
'count': len(recordings),
'recordings': recordings
})
@app.route('/api/recordings/<filename>', methods=['GET'])
def get_recording(filename):
"""获取单个录音文件"""
return send_from_directory(UPLOAD_FOLDER, filename)
@app.route('/api/recordings/<filename>', methods=['DELETE'])
def delete_recording(filename):
"""删除录音文件"""
filepath = os.path.join(UPLOAD_FOLDER, secure_filename(filename))
if os.path.exists(filepath):
os.remove(filepath)
return jsonify({'success': True, 'message': '已删除'})
return jsonify({'error': '文件不存在'}), 404
@app.route('/api/health', methods=['GET'])
def health_check():
"""健康检查"""
vosk_exists = os.path.exists(VOSK_MODEL_PATH)
return jsonify({
'status': 'ok',
'vosk_model': 'available' if vosk_exists else 'not_found',
'funasr_model': 'lazy_load', # 懒加载
'upload_folder': UPLOAD_FOLDER,
'recordings_count': len([f for f in os.listdir(UPLOAD_FOLDER) if allowed_file(f)])
})
@app.route('/api/analyze', methods=['POST'])
def analyze_text():
"""分析转写文字内容"""
try:
data = request.get_json()
if not data or 'text' not in data:
return jsonify({'success': False, 'error': '缺少 text 参数'}), 400
text = data['text'].strip()
if len(text) < 10:
return jsonify({'success': False, 'error': '文字内容太短(至少10个字符)'}), 400
# 获取分析类型,默认为 general
analysis_type = data.get('type', 'general')
# 调用大模型分析
analysis = call_llm_for_analysis(text, analysis_type)
return jsonify({
'success': True,
'analysis': analysis,
'type': analysis_type
})
except Exception as e:
print(f"分析异常: {e}")
return jsonify({'success': False, 'error': str(e)}), 500
def get_analysis_prompt(analysis_type):
"""根据分析类型返回对应的系统提示语"""
prompts = {
'meeting': """你是一位专业的会议纪要助手。请对提供的会议内容进行整理,要求:
## 会议纪要结构:
📋 会议基本信息:
- 会议主题/议题(根据内容推断)
- 参会人员(提取提到的角色)
- 会议时间(如有提及)
📌 会议决议/共识:
逐条列出会议达成的主要决议,每条包含:
- 决议事项
- 具体内容和依据
✅ 行动项(TODO):
明确列出需要执行的任务:
- 任务内容
- 负责人(如提及)
- 时间节点/截止日期(如提及)
⚠️ 待确认/风险点:
- 需要进一步明确的事项
- 提到的风险或注意事项
💬 其他重要讨论:
记录其他有价值的讨论内容
⚠️ 要求:
- 不歪曲原意
- 行动项必须具体可执行
- 使用中文回答""",
'weekly': """你是一位专业的工作周报助手。请对提供的工作内容进行周报整理,要求:
## 周报结构:
📊 本周工作完成情况:
逐条列出已完成的工作,每条包含:
- 工作内容
- 完成情况/成果
- 相关数据或指标(如有)
🔄 本周进行中工作:
- 工作内容
- 当前进度
- 预计完成时间
❓ 遇到的问题/困难:
- 问题描述
- 已尝试的解决方案
- 需要的支持或资源
📅 下周工作计划:
- 计划完成的工作项
- 优先级排序
- 预期目标
💡 其他事项:
- 心得体会
- 建议或想法
⚠️ 要求:
- 量化成果(数字、百分比等)
- 问题要具体
- 计划要可衡量
- 使用中文回答""",
'daily': """你是一位专业的工作日报助手。请对提供的工作内容进行日报整理,要求:
## 日报结构:
✅ 今日完成工作:
逐条列出今日完成的工作,按重要性排序:
- 工作事项
- 完成结果
- 耗时(如可推断)
🔄 进行中工作:
- 工作内容
- 今日进展
- 明日继续计划
❓ 今日问题/阻塞:
- 遇到的问题
- 影响程度
- 尝试的解决方案
📅 明日工作计划:
- 计划工作项
- 预期目标
- 需要协调的事项
🙋 需要支持:
- 需要领导/同事协助的事项
- 需要的资源
⚠️ 要求:
- 简洁明了
- 问题不回避
- 计划具体
- 使用中文回答""",
'general': """你是一位专业的内容分析专家。请对提供的文字内容进行深度分析和总结,要求:
## 分析原则:
1. 忠实原文:不歪曲、不遗漏原文的核心信息
2. 深入挖掘:提取隐含的观点、态度和意图
3. 结构清晰:层次分明,便于阅读理解
4. 内容充实:每个要点都要有具体支撑
## 输出格式:
📌 核心要点(3-5条):
每条要点需包含:
- 要点概括(一句话)
- 具体内容(详细说明,引用原文关键句)
🔍 深度分析:
- 说话者/作者的主要观点或立场
- 情绪态度(积极/消极/中性)
- 潜在的意图或目的
💡 关键信息提取:
- 涉及的人物、时间、地点
- 重要数据或事实
- 行动项或待办事项
📝 详细总结:
用3-5句话全面概括全文内容
⚠️ 注意事项:
- 不要添加原文没有的信息
- 不要过度解读,保持客观
- 使用中文回答"""
}
return prompts.get(analysis_type, prompts['general'])
def call_llm_for_analysis(text, analysis_type='general'):
"""调用 Moonshot 大模型进行内容分析总结"""
system_prompt = get_analysis_prompt(analysis_type)
try:
# 使用 Moonshot API
import openai
api_key = os.environ.get('MOONSHOT_API_KEY')
if not api_key:
print("Moonshot API key 未配置")
return simple_analysis(text)
# Moonshot 使用 OpenAI 兼容接口
client = openai.OpenAI(
api_key=api_key,
base_url="https://api.moonshot.cn/v1"
)
# 根据类型调整提示语
type_names = {
'meeting': '会议',
'weekly': '周报',
'daily': '日报',
'general': '内容'
}
type_name = type_names.get(analysis_type, '内容')
response = client.chat.completions.create(
model="moonshot-v1-8k", # 使用 8k 上下文模型
messages=[
{"role": "system", "content": system_prompt},
{"role": "user", "content": f"请对以下{type_name}进行整理分析:\n\n{text}"}
],
temperature=0.3,
max_tokens=2500 # 增加token数以获得更详细的回答
)
result = response.choices[0].message.content
print(f"Moonshot 分析完成,返回 {len(result)} 字符")
return result
except Exception as e:
print(f"Moonshot API 调用失败: {e}")
# 失败时回退到本地分析
return simple_analysis(text)
def simple_analysis(text):
"""本地简单分析(无需外部 API)"""
lines = text.split('\n')
sentences = [s.strip() for s in text.replace('。', '\n').replace('!', '\n').replace('?', '\n').split('\n') if s.strip()]
# 提取关键句子(长度适中,包含信息量)
key_sentences = [s for s in sentences if 20 <= len(s) <= 100][:5]
# 统计信息
char_count = len(text)
word_count = len(text.split())
result = f"""📌 核心要点:
"""
if key_sentences:
for i, sent in enumerate(key_sentences[:4], 1):
result += f"{i}. {sent}\n"
else:
result += "1. (内容较短,未提取到明显要点)\n"
result += f"""
💡 关键信息:
- 总字数:约 {char_count} 字
- 句子数:约 {len(sentences)} 句
- 内容类型:{"对话/访谈" if "?" in text or "说" in text else "叙述/说明"}
📝 简要总结:
{text[:100]}{"..." if len(text) > 100 else ""}
"""
return result.strip()
@app.route('/')
def index():
return send_from_directory('../frontend', 'index.html')
@app.route('/<path:path>')
def static_files(path):
return send_from_directory('../frontend', path)
if __name__ == '__main__':
cert_file = '/home/ubuntu/nanocode/audio-recorder/cert.pem'
key_file = '/home/ubuntu/nanocode/audio-recorder/key.pem'
print(f"🎙️ 录音服务启动中...")
print(f"📁 上传目录: {UPLOAD_FOLDER}")
print(f"🔤 Vosk 模型: {VOSK_MODEL_PATH}")
print(f"🤖 FunASR: 延迟加载(首次转写时加载,约100MB内存)")
# 检查证书
if os.path.exists(cert_file) and os.path.exists(key_file):
print(f"🔒 HTTPS 模式: https://0.0.0.0:8891")
context = ssl.SSLContext(ssl.PROTOCOL_TLS_SERVER)
context.load_cert_chain(cert_file, key_file)
app.run(
host='0.0.0.0',
port=8891,
ssl_context=context,
debug=False
)
else:
print(f"⚠️ 证书不存在,使用 HTTP: http://0.0.0.0:8890")
app.run(
host='0.0.0.0',
port=8890,
debug=False
)
+123
View File
@@ -0,0 +1,123 @@
#!/usr/bin/env python3
"""
录音后端服务 - HTTPS 版本
"""
from flask import Flask, request, jsonify, send_from_directory
from flask_cors import CORS
from werkzeug.utils import secure_filename
import os
import ssl
from datetime import datetime
app = Flask(__name__)
CORS(app)
# 配置
UPLOAD_FOLDER = '/home/ubuntu/nanocode/audio-recorder/uploads'
ALLOWED_EXTENSIONS = {'webm', 'mp4', 'wav', 'mp3', 'ogg'}
MAX_CONTENT_LENGTH = 100 * 1024 * 1024 # 100MB
app.config['MAX_CONTENT_LENGTH'] = MAX_CONTENT_LENGTH
# 确保上传目录存在
os.makedirs(UPLOAD_FOLDER, exist_ok=True)
def allowed_file(filename):
return '.' in filename and filename.rsplit('.', 1)[1].lower() in ALLOWED_EXTENSIONS
@app.route('/api/upload', methods=['POST'])
def upload_audio():
"""接收音频文件上传"""
if 'audio' not in request.files:
return jsonify({'error': '没有文件'}), 400
file = request.files['audio']
if file.filename == '':
return jsonify({'error': '文件名为空'}), 400
if file and allowed_file(file.filename):
timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
filename = f"recording_{timestamp}.webm"
filepath = os.path.join(UPLOAD_FOLDER, filename)
file.save(filepath)
file_size = os.path.getsize(filepath)
return jsonify({
'success': True,
'filename': filename,
'filepath': filepath,
'size': file_size,
'size_human': f"{file_size / 1024:.1f} KB",
'timestamp': timestamp
}), 200
return jsonify({'error': '不支持的文件类型'}), 400
@app.route('/api/recordings', methods=['GET'])
def list_recordings():
"""列出所有录音文件"""
recordings = []
for filename in sorted(os.listdir(UPLOAD_FOLDER), reverse=True):
if allowed_file(filename):
filepath = os.path.join(UPLOAD_FOLDER, filename)
stat = os.stat(filepath)
recordings.append({
'filename': filename,
'size': stat.st_size,
'size_human': f"{stat.st_size / 1024:.1f} KB",
'created': datetime.fromtimestamp(stat.st_mtime).isoformat()
})
return jsonify({
'count': len(recordings),
'recordings': recordings
})
@app.route('/api/recordings/<filename>', methods=['GET'])
def get_recording(filename):
"""获取单个录音文件"""
return send_from_directory(UPLOAD_FOLDER, filename)
@app.route('/api/health', methods=['GET'])
def health_check():
"""健康检查"""
return jsonify({
'status': 'ok',
'upload_folder': UPLOAD_FOLDER,
'recordings_count': len([f for f in os.listdir(UPLOAD_FOLDER) if allowed_file(f)])
})
@app.route('/')
def index():
return send_from_directory('../frontend', 'index.html')
@app.route('/<path:path>')
def static_files(path):
return send_from_directory('../frontend', path)
if __name__ == '__main__':
context = ssl.SSLContext(ssl.PROTOCOL_TLS_SERVER)
context.load_cert_chain('/home/ubuntu/nanocode/audio-recorder/cert.pem',
'/home/ubuntu/nanocode/audio-recorder/key.pem')
print(f"🎙️ HTTPS 录音服务启动中...")
print(f"📁 上传目录: {UPLOAD_FOLDER}")
print(f"🔒 HTTPS 地址: https://129.211.173.87:8891")
print(f"⚠️ 由于是自签名证书,浏览器会显示安全警告,请点击'高级'->'继续访问'")
app.run(
host='0.0.0.0',
port=8891,
ssl_context=context,
debug=False
)
+182
View File
@@ -0,0 +1,182 @@
#!/usr/bin/env python3
"""
录音后端服务 V2 - 集成语音转文字
"""
import warnings
warnings.filterwarnings('ignore')
from flask import Flask, request, jsonify, send_from_directory
from flask_cors import CORS
from werkzeug.utils import secure_filename
import os
import ssl
from datetime import datetime
# 导入转写模块 - 使用 Vosk
from transcribe_api import transcribe_audio
MODEL_PATH = '/home/ubuntu/nanocode/audio-recorder/models/cn-model'
app = Flask(__name__)
CORS(app)
# 配置
UPLOAD_FOLDER = '/home/ubuntu/nanocode/audio-recorder/uploads'
ALLOWED_EXTENSIONS = {'webm', 'mp4', 'wav', 'mp3', 'ogg'}
MAX_CONTENT_LENGTH = 100 * 1024 * 1024
app.config['MAX_CONTENT_LENGTH'] = MAX_CONTENT_LENGTH
os.makedirs(UPLOAD_FOLDER, exist_ok=True)
def allowed_file(filename):
return '.' in filename and filename.rsplit('.', 1)[1].lower() in ALLOWED_EXTENSIONS
@app.route('/api/upload', methods=['POST'])
def upload_audio():
"""接收音频文件上传"""
if 'audio' not in request.files:
return jsonify({'error': '没有文件'}), 400
file = request.files['audio']
if file.filename == '':
return jsonify({'error': '文件名为空'}), 400
if file and allowed_file(file.filename):
timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
filename = f"recording_{timestamp}.webm"
filepath = os.path.join(UPLOAD_FOLDER, filename)
file.save(filepath)
file_size = os.path.getsize(filepath)
return jsonify({
'success': True,
'filename': filename,
'filepath': filepath,
'size': file_size,
'size_human': f"{file_size / 1024:.1f} KB",
'timestamp': timestamp
}), 200
return jsonify({'error': '不支持的文件类型'}), 400
@app.route('/api/transcribe', methods=['POST'])
def transcribe():
"""语音转文字"""
if 'audio' not in request.files:
return jsonify({'error': '没有音频文件'}), 400
file = request.files['audio']
if file.filename == '':
return jsonify({'error': '文件名为空'}), 400
try:
# 保存上传的音频
timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
filename = f"recording_{timestamp}.webm"
filepath = os.path.join(UPLOAD_FOLDER, filename)
file.save(filepath)
print(f"🎵 收到音频: {filename}, 大小: {os.path.getsize(filepath)} bytes")
# 转写
text, error = transcribe_audio(filepath)
if error:
print(f"❌ 转写失败: {error}")
return jsonify({
'success': False,
'error': error,
'filename': filename
}), 500
print(f"✅ 转写成功: {text[:50]}...")
return jsonify({
'success': True,
'text': text or '(未识别到语音)',
'filename': filename,
'model_type': 'vosk' if os.path.exists(MODEL_PATH) else 'mock'
})
except Exception as e:
print(f"❌ 服务器错误: {e}")
return jsonify({
'success': False,
'error': str(e)
}), 500
@app.route('/api/recordings', methods=['GET'])
def list_recordings():
"""列出所有录音文件"""
recordings = []
for filename in sorted(os.listdir(UPLOAD_FOLDER), reverse=True):
if allowed_file(filename):
filepath = os.path.join(UPLOAD_FOLDER, filename)
stat = os.stat(filepath)
recordings.append({
'filename': filename,
'size': stat.st_size,
'size_human': f"{stat.st_size / 1024:.1f} KB",
'created': datetime.fromtimestamp(stat.st_mtime).isoformat()
})
return jsonify({
'count': len(recordings),
'recordings': recordings
})
@app.route('/api/recordings/<filename>', methods=['GET'])
def get_recording(filename):
"""获取单个录音文件"""
return send_from_directory(UPLOAD_FOLDER, filename)
@app.route('/api/health', methods=['GET'])
def health_check():
"""健康检查"""
model_exists = os.path.exists(MODEL_PATH)
return jsonify({
'status': 'ok',
'model_loaded': model_exists,
'model_path': MODEL_PATH,
'upload_folder': UPLOAD_FOLDER,
'recordings_count': len([f for f in os.listdir(UPLOAD_FOLDER) if allowed_file(f)])
})
@app.route('/')
def index():
return send_from_directory('../frontend', 'index.html')
@app.route('/<path:path>')
def static_files(path):
return send_from_directory('../frontend', path)
if __name__ == '__main__':
context = ssl.SSLContext(ssl.PROTOCOL_TLS_SERVER)
context.load_cert_chain(
'/home/ubuntu/nanocode/audio-recorder/cert.pem',
'/home/ubuntu/nanocode/audio-recorder/key.pem'
)
print(f"🎙️ HTTPS 录音服务 V2 启动中...")
print(f"📁 上传目录: {UPLOAD_FOLDER}")
print(f"🔤 模型路径: {MODEL_PATH}")
print(f"🔒 HTTPS 地址: https://129.211.173.87:8891")
print(f"⚠️ 提示: {'✅ 模型已安装' if os.path.exists(MODEL_PATH) else '⚠️ 使用模拟转写,请安装模型'}")
app.run(
host='0.0.0.0',
port=8891,
ssl_context=context,
debug=False
)
+300
View File
@@ -0,0 +1,300 @@
#!/usr/bin/env python3
"""
录音后端服务 - 集成语音转文字
"""
from flask import Flask, request, jsonify, send_from_directory
from flask_cors import CORS
from werkzeug.utils import secure_filename
import os
import ssl
import wave
import json
from datetime import datetime
app = Flask(__name__)
CORS(app)
# 配置
UPLOAD_FOLDER = '/home/ubuntu/nanocode/audio-recorder/uploads'
MODEL_PATH = '/home/ubuntu/nanocode/audio-recorder/models/cn-model'
ALLOWED_EXTENSIONS = {'webm', 'mp4', 'wav', 'mp3', 'ogg'}
MAX_CONTENT_LENGTH = 100 * 1024 * 1024
app.config['MAX_CONTENT_LENGTH'] = MAX_CONTENT_LENGTH
os.makedirs(UPLOAD_FOLDER, exist_ok=True)
# 全局 ASR 模型(延迟加载)
funasr_model = None
vosk_model = None
def get_funasr_model():
"""加载 FunASR 模型(阿里,中文识别效果好)- 优先使用"""
global funasr_model
if funasr_model is None:
try:
from funasr import AutoModel
print("🔄 加载 FunASR 模型(阿里达摩院,中文优先)...")
funasr_model = AutoModel(
model="paraformer-zh",
vad_model="fsmn-vad",
punc_model="ct-punc",
)
print("✅ FunASR 模型加载完成")
except Exception as e:
print(f"⚠️ FunASR 加载失败: {e}")
return None
return funasr_model
def get_vosk_model():
"""加载 Vosk 模型(备用)"""
global vosk_model
if vosk_model is None:
try:
from vosk import Model
if os.path.exists(MODEL_PATH):
print(f"🔄 加载 Vosk 模型(备用)...")
vosk_model = Model(MODEL_PATH)
print("✅ Vosk 模型加载完成")
else:
print(f"⚠️ Vosk 模型不存在: {MODEL_PATH}")
return None
except Exception as e:
print(f"⚠️ Vosk 加载失败: {e}")
return None
return vosk_model
def allowed_file(filename):
return '.' in filename and filename.rsplit('.', 1)[1].lower() in ALLOWED_EXTENSIONS
def convert_to_wav(input_path, output_path):
"""将音频转换为 WAV 格式(16kHz, 单声道)"""
import subprocess
try:
cmd = [
'ffmpeg', '-y', '-i', input_path,
'-ar', '16000', '-ac', '1', '-f', 'wav',
output_path
]
subprocess.run(cmd, check=True, capture_output=True, timeout=30)
return True
except Exception as e:
print(f"转换失败: {e}")
return False
def transcribe_with_funasr(audio_path):
"""使用 FunASR 转写(优先,中文效果好)"""
model = get_funasr_model()
if not model:
return None, "FunASR 模型未加载"
try:
res = model.generate(input=audio_path, batch_size_s=300)
text = res[0]["text"] if res and len(res) > 0 else ""
return text, None
except Exception as e:
return None, f"FunASR 错误: {str(e)}"
def transcribe_with_vosk(audio_path):
"""使用 Vosk 转写音频"""
from vosk import KaldiRecognizer
model = get_asr_model()
if not model:
return None, "模型未加载"
try:
# 转换为 WAV
wav_path = audio_path + '.wav'
if not convert_to_wav(audio_path, wav_path):
return None, "音频转换失败"
# 读取 WAV 文件
wf = wave.open(wav_path, 'rb')
if wf.getnchannels() != 1 or wf.getsampwidth() != 2 or wf.getcomptype() != 'NONE':
return None, "音频格式不支持"
# 创建识别器
rec = KaldiRecognizer(model, wf.getframerate())
rec.SetWords(True)
# 识别
results = []
while True:
data = wf.readframes(4000)
if len(data) == 0:
break
if rec.AcceptWaveform(data):
result = json.loads(rec.Result())
if result.get('text'):
results.append(result['text'])
# 获取最终结果
final_result = json.loads(rec.FinalResult())
if final_result.get('text'):
results.append(final_result['text'])
# 清理临时文件
wf.close()
if os.path.exists(wav_path):
os.remove(wav_path)
return ' '.join(results), None
except Exception as e:
return None, str(e)
@app.route('/api/upload', methods=['POST'])
def upload_audio():
"""接收音频文件上传"""
if 'audio' not in request.files:
return jsonify({'error': '没有文件'}), 400
file = request.files['audio']
if file.filename == '':
return jsonify({'error': '文件名为空'}), 400
if file and allowed_file(file.filename):
timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
filename = f"recording_{timestamp}.webm"
filepath = os.path.join(UPLOAD_FOLDER, filename)
file.save(filepath)
file_size = os.path.getsize(filepath)
return jsonify({
'success': True,
'filename': filename,
'filepath': filepath,
'size': file_size,
'size_human': f"{file_size / 1024:.1f} KB",
'timestamp': timestamp
}), 200
return jsonify({'error': '不支持的文件类型'}), 400
@app.route('/api/transcribe', methods=['POST'])
def transcribe():
"""语音转文字 - 优先使用 FunASR(中文效果好),失败回退 Vosk"""
if 'audio' not in request.files:
return jsonify({'error': '没有音频文件'}), 400
file = request.files['audio']
if file.filename == '':
return jsonify({'error': '文件名为空'}), 400
try:
# 保存上传的音频
timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')
filename = f"recording_{timestamp}.webm"
filepath = os.path.join(UPLOAD_FOLDER, filename)
file.save(filepath)
# 转换为 WAV 格式(两种引擎都需要)
wav_path = filepath + '.wav'
if not convert_to_wav(filepath, wav_path):
return jsonify({
'success': False,
'error': '音频格式转换失败'
}), 500
# 1. 优先使用 FunASR(阿里,中文识别效果好)
text, error = transcribe_with_funasr(wav_path)
engine = "funasr"
# 2. FunASR 失败,回退到 Vosk
if not text:
print(f"⚠️ FunASR 失败({error}),回退到 Vosk...")
text, error = transcribe_with_vosk(wav_path)
engine = "vosk"
# 清理临时 WAV 文件
if os.path.exists(wav_path):
os.remove(wav_path)
if error and not text:
return jsonify({
'success': False,
'error': error,
'filename': filename
}), 500
return jsonify({
'success': True,
'text': text or '(未识别到语音)',
'filename': filename,
'engine': engine
})
except Exception as e:
return jsonify({
'success': False,
'error': str(e)
}), 500
@app.route('/api/recordings', methods=['GET'])
def list_recordings():
"""列出所有录音文件"""
recordings = []
for filename in sorted(os.listdir(UPLOAD_FOLDER), reverse=True):
if allowed_file(filename):
filepath = os.path.join(UPLOAD_FOLDER, filename)
stat = os.stat(filepath)
recordings.append({
'filename': filename,
'size': stat.st_size,
'size_human': f"{stat.st_size / 1024:.1f} KB",
'created': datetime.fromtimestamp(stat.st_mtime).isoformat()
})
return jsonify({
'count': len(recordings),
'recordings': recordings
})
@app.route('/api/recordings/<filename>', methods=['GET'])
def get_recording(filename):
"""获取单个录音文件"""
return send_from_directory(UPLOAD_FOLDER, filename)
@app.route('/api/health', methods=['GET'])
def health_check():
"""健康检查"""
model_exists = os.path.exists(MODEL_PATH)
return jsonify({
'status': 'ok',
'model_loaded': model_exists,
'model_path': MODEL_PATH,
'upload_folder': UPLOAD_FOLDER,
'recordings_count': len([f for f in os.listdir(UPLOAD_FOLDER) if allowed_file(f)])
})
@app.route('/')
def index():
return send_from_directory('../frontend', 'index.html')
@app.route('/<path:path>')
def static_files(path):
return send_from_directory('../frontend', path)
if __name__ == '__main__':
context = ssl.SSLContext(ssl.PROTOCOL_TLS_SERVER)
context.load_cert_chain(
'/home/ubuntu/nanocode/audio-recorder/cert.pem',
'/home/ubuntu/nanocode/audio-recorder/key.pem'
)
print(f"🎙️ HTTPS 录音服务启动中...")
print(f"📁 上传目录: {UPLOAD_FOLDER}")
print(f"🔤 模型路径: {MODEL_PATH}")
print(f"🔒 HTTPS 地址: https://129.211.173.87:8891")
if not os.path.exists(MODEL_PATH):
print(f"⚠️ 警告: ASR 模型未找到,请先下载模型")
app.run(
host='0.0.0.0',
port=8891,
ssl_context=context,
debug=False
)
+4
View File
@@ -0,0 +1,4 @@
Flask==3.0.0
Flask-CORS==4.0.0
Werkzeug==3.0.1
gunicorn==21.2.0
+146
View File
@@ -0,0 +1,146 @@
#!/usr/bin/env python3
"""
语音转文字 - 使用在线 API 方案
支持:
1. Vosk 本地模型(如果有)
2. Whisper API(OpenAI/DeepSeek)
3. 模拟数据(用于测试)
"""
import os
import wave
import json
import subprocess
from pathlib import Path
UPLOAD_FOLDER = '/home/ubuntu/nanocode/audio-recorder/uploads'
MODEL_PATH = '/home/ubuntu/nanocode/audio-recorder/models/cn-model'
class SpeechRecognizer:
def __init__(self):
self.model = None
self.model_type = self._detect_model()
def _detect_model(self):
"""检测可用的 ASR 模型"""
if os.path.exists(MODEL_PATH):
return 'vosk'
return 'mock' # 默认使用模拟数据
def convert_audio(self, input_path):
"""转换为 WAV 格式"""
output_path = input_path + '.wav'
try:
cmd = [
'ffmpeg', '-y', '-i', input_path,
'-ar', '16000', '-ac', '1', '-f', 'wav',
output_path
]
result = subprocess.run(cmd, capture_output=True, text=True, timeout=30)
if result.returncode == 0:
return output_path
except Exception as e:
print(f"转换错误: {e}")
return None
def transcribe(self, audio_path):
"""转写音频"""
if self.model_type == 'vosk':
return self._transcribe_vosk(audio_path)
else:
return self._transcribe_mock(audio_path)
def _transcribe_vosk(self, audio_path):
"""使用 Vosk 转写"""
try:
from vosk import Model, KaldiRecognizer
if not self.model:
self.model = Model(MODEL_PATH)
# 转换为 WAV
wav_path = self.convert_audio(audio_path)
if not wav_path:
return None, "音频转换失败"
# 读取音频
wf = wave.open(wav_path, 'rb')
if wf.getnchannels() != 1 or wf.getsampwidth() != 2:
wf.close()
os.remove(wav_path)
return None, "音频格式不支持"
# 识别
rec = KaldiRecognizer(self.model, wf.getframerate())
rec.SetWords(True)
results = []
while True:
data = wf.readframes(4000)
if len(data) == 0:
break
if rec.AcceptWaveform(data):
result = json.loads(rec.Result())
if result.get('text'):
results.append(result['text'])
final = json.loads(rec.FinalResult())
if final.get('text'):
results.append(final['text'])
wf.close()
os.remove(wav_path)
return ' '.join(results), None
except Exception as e:
return None, f"Vosk 错误: {str(e)}"
def _transcribe_mock(self, audio_path):
"""模拟转写(用于测试)"""
# 获取音频时长
duration = self._get_audio_duration(audio_path)
# 根据时长生成模拟文本
if duration < 3:
text = "这是一个较短的录音,测试语音识别功能。"
elif duration < 10:
text = """这是一个测试录音。
语音识别功能正在开发中。
目前使用模拟数据展示界面效果。
请在服务器上安装 Vosk 模型以获得真实转写功能。"""
else:
text = """这是一个较长的录音测试。
当前使用的是模拟转写功能。
真实转写需要安装 Vosk 中文模型:
1. cd /home/ubuntu/nanocode/audio-recorder/models
2. wget https://alphacephei.com/vosk/models/vosk-model-small-cn-0.22.zip
3. unzip vosk-model-small-cn-0.22.zip
4. mv vosk-model-small-cn-0.22 cn-model
5. 重启后端服务
或者使用在线 API 方案(DeepSeek/Whisper)。"""
return text, None
def _get_audio_duration(self, audio_path):
"""获取音频时长(秒)"""
try:
cmd = ['ffprobe', '-v', 'error', '-show_entries',
'format=duration', '-of', 'default=noprint_wrappers=1:nokey=1', audio_path]
result = subprocess.run(cmd, capture_output=True, text=True, timeout=5)
return float(result.stdout.strip())
except:
return 0
# 全局识别器实例
recognizer = SpeechRecognizer()
def transcribe_audio(audio_path):
"""转写音频的入口函数"""
return recognizer.transcribe(audio_path)
+142
View File
@@ -0,0 +1,142 @@
#!/usr/bin/env python3
"""
语音转文字 - FunASR 版本(阿里达摩院)
支持中文识别、自动标点
"""
import os
import subprocess
import warnings
warnings.filterwarnings('ignore')
UPLOAD_FOLDER = '/home/ubuntu/nanocode/audio-recorder/uploads'
MODEL_CACHE = os.path.expanduser('~/.cache/modelscope')
class FunASRRecognizer:
def __init__(self):
self.model = None
self.model_type = 'funasr' if self._check_model_cached() else 'vosk'
def _check_model_cached(self):
"""检查 FunASR 模型是否已下载"""
# 检查常用的模型路径
possible_paths = [
os.path.join(MODEL_CACHE, 'hub', 'damo', 'speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch'),
os.path.join(MODEL_CACHE, 'hub', 'iic', 'speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch'),
]
return any(os.path.exists(p) for p in possible_paths)
def _load_model(self):
"""加载 FunASR 模型"""
if self.model is None:
try:
from funasr import AutoModel
print("🔄 加载 FunASR 模型...")
self.model = AutoModel(
model="paraformer-zh",
vad_model="fsmn-vad",
punc_model="ct-punc",
)
print("✅ FunASR 模型加载完成")
self.model_type = 'funasr'
except Exception as e:
print(f"⚠️ FunASR 加载失败: {e}, 回退到 Vosk")
self.model_type = 'vosk'
raise
return self.model
def convert_to_wav(self, input_path):
"""转换为 WAV 格式"""
output_path = input_path + '.wav'
try:
cmd = [
'ffmpeg', '-y', '-i', input_path,
'-ar', '16000', '-ac', '1', '-f', 'wav',
output_path
]
result = subprocess.run(cmd, capture_output=True, timeout=30)
if result.returncode == 0 and os.path.exists(output_path):
return output_path
except Exception as e:
print(f"转换错误: {e}")
return None
def transcribe(self, audio_path):
"""转写音频"""
# 先转换为 wav
wav_path = self.convert_to_wav(audio_path)
if not wav_path:
return None, "音频转换失败"
try:
# 尝试使用 FunASR
model = self._load_model()
if self.model_type == 'funasr':
# FunASR 识别
res = model.generate(input=wav_path, batch_size_s=300)
text = res[0]["text"]
else:
# 回退到 Vosk
text, error = self._transcribe_vosk(wav_path)
if error:
raise Exception(error)
# 清理临时文件
if os.path.exists(wav_path):
os.remove(wav_path)
return text, None
except Exception as e:
# 清理临时文件
if os.path.exists(wav_path):
os.remove(wav_path)
return None, str(e)
def _transcribe_vosk(self, wav_path):
"""备用:Vosk 识别"""
try:
from vosk import Model, KaldiRecognizer
import json
import wave
model_path = '/home/ubuntu/nanocode/audio-recorder/models/cn-model'
model = Model(model_path)
wf = wave.open(wav_path, 'rb')
rec = KaldiRecognizer(model, wf.getframerate())
results = []
while True:
data = wf.readframes(4000)
if len(data) == 0:
break
if rec.AcceptWaveform(data):
result = json.loads(rec.Result())
if result.get('text'):
results.append(result['text'])
final = json.loads(rec.FinalResult())
if final.get('text'):
results.append(final['text'])
wf.close()
return ' '.join(results), None
except Exception as e:
return None, f"Vosk 错误: {str(e)}"
# 全局识别器
_recognizer = None
def get_recognizer():
global _recognizer
if _recognizer is None:
_recognizer = FunASRRecognizer()
return _recognizer
def transcribe_audio(audio_path):
"""转写音频入口"""
recognizer = get_recognizer()
return recognizer.transcribe(audio_path)
+19
View File
@@ -0,0 +1,19 @@
#!/bin/bash
# 下载 Vosk 中文模型
cd /home/ubuntu/nanocode/audio-recorder/models
echo "📥 下载 Vosk 中文模型..."
wget -O vosk-model-small-cn-0.22.zip \
https://alphacephei.com/vosk/models/vosk-model-small-cn-0.22.zip
echo "📦 解压模型..."
unzip vosk-model-small-cn-0.22.zip
mv vosk-model-small-cn-0.22 cn-model
echo "🗑️ 清理..."
rm vosk-model-small-cn-0.22.zip
echo "✅ 模型安装完成!"
echo "📍 路径: /home/ubuntu/nanocode/audio-recorder/models/cn-model"
echo "🔄 请重启服务: pkill -f app_with_asr.py && python3 backend/app_with_asr.py"
+730
View File
@@ -0,0 +1,730 @@
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=no">
<title>录音工具</title>
<style>
* {
margin: 0;
padding: 0;
box-sizing: border-box;
}
body {
font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
min-height: 100vh;
display: flex;
flex-direction: column;
align-items: center;
padding: 15px;
}
.container {
width: 100%;
max-width: 400px;
background: white;
border-radius: 16px;
padding: 20px;
box-shadow: 0 15px 50px rgba(0,0,0,0.3);
}
h1 {
text-align: center;
color: #333;
margin-bottom: 15px;
font-size: 22px;
}
.status {
text-align: center;
padding: 8px 12px;
margin-bottom: 10px;
border-radius: 8px;
font-size: 14px;
font-weight: 500;
transition: all 0.3s;
min-height: 32px;
display: flex;
align-items: center;
justify-content: center;
}
.status.idle {
background: #f0f0f0;
color: #666;
}
.status.recording {
background: #ffebee;
color: #c62828;
animation: pulse 1.5s infinite;
}
.status.playing {
background: #e3f2fd;
color: #1565c0;
}
.status.transcribing {
background: #fff3e0;
color: #e65100;
}
@keyframes pulse {
0%, 100% { opacity: 1; }
50% { opacity: 0.7; }
}
.timer {
text-align: center;
font-size: 36px;
font-weight: 300;
color: #333;
margin: 8px 0;
font-variant-numeric: tabular-nums;
line-height: 1;
}
.button-row {
display: flex;
gap: 10px;
margin-bottom: 12px;
}
button {
flex: 1;
padding: 14px 16px;
border: none;
border-radius: 10px;
font-size: 16px;
font-weight: 600;
cursor: pointer;
transition: all 0.3s;
display: flex;
align-items: center;
justify-content: center;
gap: 6px;
}
button:disabled {
opacity: 0.5;
cursor: not-allowed;
}
button:active:not(:disabled) {
transform: scale(0.98);
}
.btn-start {
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
color: white;
}
.btn-stop {
background: #ef5350;
color: white;
}
.btn-play {
background: #42a5f5;
color: white;
flex: 0.6;
}
.progress-info {
flex: 0.4;
display: flex;
align-items: center;
justify-content: center;
background: #f5f5f5;
border-radius: 10px;
font-size: 14px;
color: #666;
}
.audio-player {
margin-top: 12px;
padding: 10px;
background: #f5f5f5;
border-radius: 10px;
display: none;
}
.audio-player.show {
display: block;
}
audio {
width: 100%;
height: 36px;
}
.icon {
font-size: 18px;
}
.error {
background: #ffebee;
color: #c62828;
padding: 10px 12px;
border-radius: 8px;
margin-bottom: 12px;
font-size: 13px;
display: none;
line-height: 1.5;
}
.error.show {
display: block;
}
/* 转写结果区域 */
.transcript-box {
margin-top: 15px;
padding: 15px;
background: #f8f9fa;
border-radius: 12px;
border-left: 4px solid #667eea;
display: none;
}
.transcript-box.show {
display: block;
}
.transcript-header {
display: flex;
justify-content: space-between;
align-items: center;
margin-bottom: 10px;
}
.transcript-title {
font-weight: 600;
color: #333;
font-size: 15px;
}
.transcript-actions {
display: flex;
gap: 8px;
}
.btn-small {
padding: 6px 12px;
font-size: 12px;
border-radius: 6px;
border: none;
cursor: pointer;
background: #e3f2fd;
color: #1565c0;
}
.transcript-content {
background: white;
padding: 12px;
border-radius: 8px;
min-height: 80px;
font-size: 15px;
line-height: 1.6;
color: #333;
white-space: pre-wrap;
word-wrap: break-word;
}
.transcript-content:empty::before {
content: '转写中...';
color: #999;
}
/* AI 改写按钮区域 */
.ai-actions {
margin-top: 12px;
display: flex;
gap: 8px;
flex-wrap: wrap;
}
.btn-ai {
padding: 8px 14px;
font-size: 13px;
border-radius: 8px;
border: none;
cursor: pointer;
background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
color: white;
font-weight: 500;
transition: all 0.3s;
}
.btn-ai:hover:not(:disabled) {
transform: translateY(-1px);
box-shadow: 0 4px 12px rgba(102, 126, 234, 0.4);
}
.btn-ai:disabled {
opacity: 0.6;
cursor: not-allowed;
}
/* AI 改写结果区域 */
.analysis-box {
margin-top: 15px;
padding: 15px;
background: #f0f7ff;
border-radius: 12px;
border-left: 4px solid #42a5f5;
display: none;
}
.analysis-box.show {
display: block;
}
.analysis-header {
display: flex;
justify-content: space-between;
align-items: center;
margin-bottom: 10px;
}
.analysis-title {
font-weight: 600;
color: #1565c0;
font-size: 15px;
}
.analysis-actions {
display: flex;
align-items: center;
gap: 8px;
}
.analysis-type {
font-size: 12px;
color: #666;
background: white;
padding: 2px 8px;
border-radius: 4px;
}
.analysis-content {
background: white;
padding: 12px;
border-radius: 8px;
min-height: 100px;
font-size: 14px;
line-height: 1.7;
color: #333;
white-space: pre-wrap;
word-wrap: break-word;
}
.analysis-loading {
text-align: center;
color: #666;
padding: 20px;
}
.analysis-loading::after {
content: '';
display: inline-block;
width: 16px;
height: 16px;
border: 2px solid #667eea;
border-top-color: transparent;
border-radius: 50%;
animation: spin 1s linear infinite;
margin-left: 8px;
vertical-align: middle;
}
@keyframes spin {
to { transform: rotate(360deg); }
}
.progress-bar {
width: 100%;
height: 4px;
background: #e0e0e0;
border-radius: 2px;
margin-top: 8px;
overflow: hidden;
display: none;
}
.progress-bar.show {
display: block;
}
.progress-fill {
height: 100%;
background: linear-gradient(90deg, #667eea, #764ba2);
width: 0%;
transition: width 0.3s;
border-radius: 2px;
}
</style>
</head>
<body>
<div class="container">
<h1>🎙️ 录音工具</h1>
<div class="error" id="errorBox"></div>
<div class="status idle" id="statusBox">准备录音</div>
<div class="timer" id="timer">00:00</div>
<!-- 第一行:开始 + 结束 -->
<div class="button-row">
<button class="btn-start" id="btnStart" onclick="startRecording()">
<span class="icon">🔴</span>开始
</button>
<button class="btn-stop" id="btnStop" onclick="stopRecording()" disabled>
<span class="icon">⏹️</span>结束
</button>
</div>
<!-- 第二行:回放 + 进度 -->
<div class="button-row">
<button class="btn-play" id="btnPlay" onclick="playRecording()" disabled>
<span class="icon">▶️</span>回放
</button>
<div class="progress-info" id="playProgress">00:00</div>
</div>
<!-- 转写进度条 -->
<div class="progress-bar" id="transcribeProgress">
<div class="progress-fill" id="progressFill"></div>
</div>
<div class="audio-player" id="audioPlayer">
<audio id="audioElement" controls></audio>
</div>
<!-- 转写结果 -->
<div class="transcript-box" id="transcriptBox">
<div class="transcript-header">
<span class="transcript-title">📝 转写结果</span>
<div class="transcript-actions">
<button class="btn-small" onclick="copyTranscript()">复制</button>
</div>
</div>
<div class="transcript-content" id="transcriptContent"></div>
<!-- AI 改写按钮 -->
<div class="ai-actions" id="aiActions" style="display: none;">
<button class="btn-ai" onclick="analyzeText('daily')">📋 日报总结</button>
<button class="btn-ai" onclick="analyzeText('weekly')">📊 周报总结</button>
<button class="btn-ai" onclick="analyzeText('general')">📝 日常总结</button>
</div>
</div>
<!-- AI 改写结果 -->
<div class="analysis-box" id="analysisBox">
<div class="analysis-header">
<span class="analysis-title">🤖 AI 改写结果</span>
<div class="analysis-actions">
<span class="analysis-type" id="analysisType"></span>
<button class="btn-small" onclick="copyAnalysis()">复制</button>
</div>
</div>
<div class="analysis-content" id="analysisContent"></div>
</div>
</div>
<script>
let mediaRecorder = null;
let audioChunks = [];
let audioBlob = null;
let audioUrl = null;
let startTime = null;
let timerInterval = null;
let currentRecordingFilename = null;
const btnStart = document.getElementById('btnStart');
const btnStop = document.getElementById('btnStop');
const btnPlay = document.getElementById('btnPlay');
const statusBox = document.getElementById('statusBox');
const timer = document.getElementById('timer');
const audioPlayer = document.getElementById('audioPlayer');
const audioElement = document.getElementById('audioElement');
const playProgress = document.getElementById('playProgress');
const errorBox = document.getElementById('errorBox');
const transcriptBox = document.getElementById('transcriptBox');
const transcriptContent = document.getElementById('transcriptContent');
const transcribeProgress = document.getElementById('transcribeProgress');
const progressFill = document.getElementById('progressFill');
const aiActions = document.getElementById('aiActions');
const analysisBox = document.getElementById('analysisBox');
const analysisContent = document.getElementById('analysisContent');
const analysisType = document.getElementById('analysisType');
function showError(msg) {
errorBox.textContent = msg;
errorBox.classList.add('show');
setTimeout(() => errorBox.classList.remove('show'), 5000);
}
function updateStatus(status, text) {
statusBox.className = 'status ' + status;
statusBox.textContent = text;
}
function formatTime(seconds) {
const mins = Math.floor(seconds / 60).toString().padStart(2, '0');
const secs = (seconds % 60).toString().padStart(2, '0');
return `${mins}:${secs}`;
}
function updateTimer() {
const elapsed = Math.floor((Date.now() - startTime) / 1000);
timer.textContent = formatTime(elapsed);
}
function updatePlayProgress() {
if (audioElement.duration) {
const current = Math.floor(audioElement.currentTime);
const total = Math.floor(audioElement.duration);
playProgress.textContent = `${formatTime(current)}/${formatTime(total)}`;
}
}
audioElement.ontimeupdate = updatePlayProgress;
audioElement.onloadedmetadata = updatePlayProgress;
async function startRecording() {
try {
const stream = await navigator.mediaDevices.getUserMedia({
audio: {
echoCancellation: true,
noiseSuppression: true,
sampleRate: 16000
}
});
const mimeType = MediaRecorder.isTypeSupported('audio/webm')
? 'audio/webm;codecs=opus'
: 'audio/mp4';
mediaRecorder = new MediaRecorder(stream, { mimeType });
audioChunks = [];
mediaRecorder.ondataavailable = (event) => {
if (event.data.size > 0) {
audioChunks.push(event.data);
}
};
mediaRecorder.onstop = () => {
audioBlob = new Blob(audioChunks, { type: mimeType });
audioUrl = URL.createObjectURL(audioBlob);
audioElement.src = audioUrl;
audioPlayer.classList.add('show');
btnPlay.disabled = false;
// 自动开始转写
transcribeAudio(audioBlob);
stream.getTracks().forEach(track => track.stop());
};
mediaRecorder.start(100);
startTime = Date.now();
timerInterval = setInterval(updateTimer, 1000);
btnStart.disabled = true;
btnStop.disabled = false;
btnPlay.disabled = true;
updateStatus('recording', '🔴 正在录音...');
audioPlayer.classList.remove('show');
transcriptBox.classList.remove('show');
aiActions.style.display = 'none';
analysisBox.classList.remove('show');
transcriptContent.textContent = '';
playProgress.textContent = '00:00';
} catch (err) {
console.error('录音失败:', err);
showError('无法访问麦克风,请检查权限设置');
}
}
function stopRecording() {
if (mediaRecorder && mediaRecorder.state !== 'inactive') {
mediaRecorder.stop();
clearInterval(timerInterval);
btnStart.disabled = false;
btnStop.disabled = true;
updateStatus('idle', '录音完成,正在转写...');
}
}
function playRecording() {
if (audioUrl) {
audioElement.play();
updateStatus('playing', '▶️ 正在播放...');
audioElement.onended = () => {
updateStatus('idle', '播放完成');
};
}
}
async function transcribeAudio(blob) {
transcribeProgress.classList.add('show');
progressFill.style.width = '30%';
try {
const formData = new FormData();
formData.append('audio', blob, `recording_${Date.now()}.webm`);
progressFill.style.width = '60%';
const response = await fetch('/api/transcribe', {
method: 'POST',
body: formData
});
progressFill.style.width = '90%';
if (response.ok) {
const result = await response.json();
progressFill.style.width = '100%';
setTimeout(() => {
transcribeProgress.classList.remove('show');
progressFill.style.width = '0%';
}, 500);
if (result.success) {
transcriptContent.textContent = result.text;
transcriptBox.classList.add('show');
aiActions.style.display = 'flex';
currentRecordingFilename = result.filename;
updateStatus('idle', '转写完成');
} else {
showError('转写失败: ' + result.error);
updateStatus('idle', '转写失败');
}
} else {
throw new Error('请求失败');
}
} catch (err) {
console.error('转写错误:', err);
transcribeProgress.classList.remove('show');
showError('转写服务暂时不可用');
updateStatus('idle', '转写失败');
}
}
function copyTranscript() {
const text = transcriptContent.textContent;
if (text) {
navigator.clipboard.writeText(text).then(() => {
const btn = event.target;
const original = btn.textContent;
btn.textContent = '已复制!';
setTimeout(() => btn.textContent = original, 1500);
});
}
}
// 复制AI分析结果
function copyAnalysis() {
const text = analysisContent.textContent;
if (text && !text.includes('正在改写中')) {
navigator.clipboard.writeText(text).then(() => {
const btn = event.target;
const original = btn.textContent;
btn.textContent = '已复制!';
setTimeout(() => btn.textContent = original, 1500);
});
}
}
// AI 分析文本
async function analyzeText(type) {
const text = transcriptContent.textContent;
if (!text || text.trim().length < 10) {
showError('转写内容太短,无法分析(至少10个字符)');
return;
}
const typeNames = {
'daily': '日报',
'weekly': '周报',
'general': '日常总结'
};
// 显示加载状态
analysisBox.classList.add('show');
analysisContent.innerHTML = '<div class="analysis-loading">AI 正在改写中</div>';
analysisType.textContent = typeNames[type];
// 禁用按钮
document.querySelectorAll('.btn-ai').forEach(btn => btn.disabled = true);
try {
const response = await fetch('/api/analyze', {
method: 'POST',
headers: {
'Content-Type': 'application/json'
},
body: JSON.stringify({
text: text,
type: type
})
});
if (response.ok) {
const result = await response.json();
if (result.success) {
analysisContent.textContent = result.analysis;
} else {
analysisContent.textContent = '改写失败: ' + result.error;
}
} else {
throw new Error('请求失败');
}
} catch (err) {
console.error('AI 分析错误:', err);
analysisContent.textContent = 'AI 服务暂时不可用,请稍后重试';
} finally {
// 启用按钮
document.querySelectorAll('.btn-ai').forEach(btn => btn.disabled = false);
}
}
// 页面加载检查
window.onload = () => {
const checks = [];
if (window.location.protocol !== 'https:' && window.location.hostname !== 'localhost') {
checks.push('❌ 需要 HTTPS 环境');
}
if (!navigator.mediaDevices || !navigator.mediaDevices.getUserMedia) {
checks.push('❌ 不支持 getUserMedia');
}
if (!window.MediaRecorder) {
checks.push('❌ 不支持 MediaRecorder');
}
if (checks.length > 0) {
let errorMsg = '环境不满足:\n' + checks.join('\n');
if (/MicroMessenger|QQ/i.test(navigator.userAgent)) {
errorMsg += '\n\n请在 Safari/Chrome 中打开';
}
showError(errorMsg);
btnStart.disabled = true;
}
};
</script>
</body>
</html>
Binary file not shown.
+34
View File
@@ -0,0 +1,34 @@
#!/bin/bash
# 启动录音服务 + HTTPS 隧道
echo "🎙️ 启动录音服务..."
# 创建上传目录
mkdir -p /home/ubuntu/nanocode/audio-recorder/uploads
cd /home/ubuntu/nanocode/audio-recorder/backend
# 检查并启动 Flask 服务
if ! pgrep -f "python3 app.py" > /dev/null; then
echo "📡 启动 Flask 服务..."
nohup python3 app.py > /tmp/audio-recorder.log 2>&1 &
sleep 2
fi
echo "✅ Flask 服务已启动: http://localhost:8890"
echo ""
# 启动 HTTPS 隧道
echo "🔒 启动 HTTPS 隧道..."
echo " 请等待 5-10 秒生成安全链接..."
echo ""
npx -y localtunnel --port 8890 &
sleep 5
echo ""
echo "📱 手机浏览器访问上面的 HTTPS 链接即可录音"
echo "⚠️ 注意:首次访问 localtunnel 需要点击 'Click to Continue'"
echo ""
echo "🛑 停止服务: pkill -f 'python3 app.py' && pkill -f 'localtunnel'"
+26
View File
@@ -0,0 +1,26 @@
#!/bin/bash
# 录音服务启动脚本
echo "🎙️ 启动录音服务..."
# 创建上传目录
mkdir -p /home/ubuntu/nanocode/audio-recorder/uploads
# 进入后端目录
cd /home/ubuntu/nanocode/audio-recorder/backend
# 检查依赖
if ! python3 -c "import flask" 2>/dev/null; then
echo "📦 安装依赖..."
pip3 install -r requirements.txt -q
fi
# 启动服务
echo "🌐 服务启动在 http://0.0.0.0:8890"
echo "📱 手机浏览器访问: http://$(hostname -I | awk '{print $1}'):8890"
echo ""
echo "⚠️ 注意:手机浏览器需要 HTTPS 才能录音"
echo " 建议用 nginx 反向代理 + SSL 证书"
echo ""
python3 app.py
+46
View File
@@ -0,0 +1,46 @@
#!/usr/bin/env python3
import warnings
warnings.filterwarnings('ignore')
from funasr import AutoModel
import os
import subprocess
print("🔄 加载 FunASR 模型...")
print(" 首次会自动下载模型(约100MB)到 ~/.cache/modelscope")
print(" 请等待...\n")
model = AutoModel(
model="paraformer-zh",
vad_model="fsmn-vad",
punc_model="ct-punc",
)
print("\n✅ FunASR 模型加载完成!")
# 测试文件
uploads_dir = 'uploads'
if os.path.exists(uploads_dir):
files = [f for f in os.listdir(uploads_dir) if f.endswith(('.webm', '.wav', '.mp3'))]
if files:
test_file = os.path.join(uploads_dir, files[0])
print(f"\n🎵 测试文件: {test_file}")
# 转换为 wav
wav_file = test_file + '.wav'
subprocess.run(['ffmpeg', '-y', '-i', test_file, '-ar', '16000', '-ac', '1', wav_file],
capture_output=True)
print("📝 开始识别...")
res = model.generate(input=wav_file, batch_size_s=300)
text = res[0]["text"]
print(f"\n✅ 识别结果:")
print(f" {text}")
if os.path.exists(wav_file):
os.remove(wav_file)
else:
print("\n⚠️ 没有测试音频")
else:
print("\n⚠️ uploads 目录不存在")