更新配置文件和Web界面，支持自定义短剧混剪片段

- 将项目版本更新至0.5.3
- 修改视觉大模型提供商为qwenvl，并更新相关API密钥和模型名称
- 新增DeepSeek API支持，更新文本生成模型设置
- 在Web界面中添加短视频生成模式的选项，调整脚本设置以支持自定义片段数量
- 修改代理设置的默认值为False

config.example.toml

@@ -1,20 +1,19 @@
 [app]
-    project_version="0.5.2"
+    project_version="0.5.3"
     # 支持视频理解的大模型提供商
     #   gemini
-    #   NarratoAPI
-    #   qwen2-vl (待增加)
-    vision_llm_provider="gemini"
+    #   qwenvl
+    vision_llm_provider="qwenvl"
     vision_analysis_prompt = "你是资深视频内容分析专家，擅长分析视频画面信息，分析下面视频画面内容，只输出客观的画面描述不要给任何总结或评价"
 
     ########## Vision Gemini API Key
     vision_gemini_api_key = ""
-    vision_gemini_model_name = "gemini-1.5-flash"
+    vision_gemini_model_name = "gemini-2.0-flash"
 
-    ########## Vision Qwen API Key
+    ########## Vision Qwen API Key (默认使用“硅基流动”的QwenVL模型)
     vision_qwenvl_api_key = ""
-    vision_qwenvl_model_name = "qwen-vl-max-latest"
-    vision_qwenvl_base_url = "https://dashscope.aliyuncs.com/compatible-mode/v1"
+    vision_qwenvl_model_name = "Qwen/Qwen2.5-VL-32B-Instruct"
+    vision_qwenvl_base_url = "https://api.siliconflow.cn/v1"
 
     ########### Vision NarratoAPI Key
     narrato_api_key = "ggyY91BAO-_ULvAqKum3XexcyN1G3dP86DEzvjZDcrg"
@@ -26,13 +25,11 @@
 
     # 用于生成文案的大模型支持的提供商 (Supported providers):
     #   openai (默认)
+    #   deepseek (默认使用“硅基流动”的模型)
     #   moonshot (月之暗面)
-    #   oneapi
-    #   g4f
-    #   azure
     #   qwen (通义千问)
     #   gemini
-    text_llm_provider="openai"
+    text_llm_provider="deepseek"
 
     ########## OpenAI API Key
     # Get your API key at https://platform.openai.com/api-keys
@@ -40,6 +37,12 @@
     text_openai_base_url = "https://api.openai.com/v1"
     text_openai_model_name = "gpt-4o-mini"
 
+    ########## DeepSeek API Key
+    # 使用 硅基流动 第三方 API Key，使用手机号注册：https://cloud.siliconflow.cn/i/pyOKqFCV
+    text_deepseek_api_key = ""
+    text_deepseek_base_url = "https://api.siliconflow.cn/v1"
+    text_deepseek_model_name = "deepseek-ai/DeepSeek-V3"
+
     ########## Moonshot API Key
     # Visit https://platform.moonshot.cn/console/api-keys to get your API key.
     text_moonshot_api_key=""
@@ -72,11 +75,6 @@
     text_qwen_model_name = "qwen-plus-1127"
     text_qwen_base_url = "https://dashscope.aliyuncs.com/compatible-mode/v1"
 
-    ########## DeepSeek API Key
-    # 使用 硅基流动 第三方 API Key，使用手机号注册：https://cloud.siliconflow.cn/i/pyOKqFCV
-    text_deepseek_api_key = ""
-    text_deepseek_base_url = "https://api.siliconflow.cn/v1"
-    text_deepseek_model_name = "deepseek-ai/DeepSeek-V3"
 
     # 字幕提供商、可选，支持 whisper 和 faster-whisper-large-v2"whisper"
     # 默认为 faster-whisper-large-v2 模型地址：https://huggingface.co/guillaumekln/faster-whisper-large-v2

webui.txt

@@ -48,6 +48,8 @@ pause
 rem set HF_ENDPOINT=https://hf-mirror.com
 streamlit run webui.py --browser.serverAddress="127.0.0.1" --server.enableCORS=True  --server.maxUploadSize=2048 --browser.gatherUsageStats=False
 
+streamlit run webui.py --server.maxUploadSize=2048
+
 请求0：
 curl -X 'POST' \
   'http://127.0.0.1:8080/api/v2/youtube/download' \

webui/components/basic_settings.py

@@ -53,7 +53,7 @@ def render_language_settings(tr):
 def render_proxy_settings(tr):
     """渲染代理设置"""
     # 获取当前代理状态
-    proxy_enabled = config.proxy.get("enabled", True)
+    proxy_enabled = config.proxy.get("enabled", False)
     proxy_url_http = config.proxy.get("http")
     proxy_url_https = config.proxy.get("https")
 

webui/components/script_settings.py

@@ -25,8 +25,17 @@ def render_script_panel(tr):
         # 渲染视频文件选择
         render_video_file(tr, params)
 
-        # 渲染视频主题和提示词
-        render_video_details(tr)
+        # 获取当前选择的脚本类型
+        script_path = st.session_state.get('video_clip_json_path', '')
+        
+        # 根据脚本类型显示不同的布局
+        if script_path == "short":
+            # Short Generate模式下显示的内容
+            render_short_generate_options(tr)
+        else:
+            # 其他模式下保持原有布局
+            # 渲染视频主题和提示词
+            render_video_details(tr)
 
         # 渲染脚本操作按钮
         render_script_buttons(tr, params)
@@ -166,6 +175,23 @@ def render_video_file(tr, params):
                 st.rerun()
 
 
+def render_short_generate_options(tr):
+    """
+    渲染Short Generate模式下的特殊选项
+    在Short Generate模式下，替换原有的输入框为自定义片段选项
+    """
+    # 显示自定义片段数量选择器
+    custom_clips = st.number_input(
+        tr("自定义片段"),
+        min_value=1,
+        max_value=20,
+        value=st.session_state.get('custom_clips', 5),
+        help=tr("设置需要生成的短视频片段数量"),
+        key="custom_clips_input"
+    )
+    st.session_state['custom_clips'] = custom_clips
+
+
 def render_video_details(tr):
     """渲染视频主题和提示词"""
     video_theme = st.text_input(tr("Video Theme"))
@@ -182,42 +208,46 @@ def render_video_details(tr):
 
 def render_script_buttons(tr, params):
     """渲染脚本操作按钮"""
-    # 新增三个输入框，放在同一行
-    input_cols = st.columns(3)
+    # 获取当前选择的脚本类型
+    script_path = st.session_state.get('video_clip_json_path', '')
     
-    with input_cols[0]:
-        skip_seconds = st.number_input(
-            "skip_seconds",
-            min_value=0,
-            value=st.session_state.get('skip_seconds', config.frames.get('skip_seconds', 0)),
-            help=tr("Skip the first few seconds"),
-            key="skip_seconds_input"
-        )
-        st.session_state['skip_seconds'] = skip_seconds
+    # 根据脚本类型显示不同的设置
+    if script_path != "short":
+        # 非短视频模式下显示原有的三个输入框
+        input_cols = st.columns(3)
         
-    with input_cols[1]:
-        threshold = st.number_input(
-            "threshold",
-            min_value=0,
-            value=st.session_state.get('threshold', config.frames.get('threshold', 30)),
-            help=tr("Difference threshold"),
-            key="threshold_input"
-        )
-        st.session_state['threshold'] = threshold
-        
-    with input_cols[2]:
-        vision_batch_size = st.number_input(
-            "vision_batch_size",
-            min_value=1,
-            max_value=20,
-            value=st.session_state.get('vision_batch_size', config.frames.get('vision_batch_size', 5)),
-            help=tr("Vision processing batch size"),
-            key="vision_batch_size_input"
-        )
-        st.session_state['vision_batch_size'] = vision_batch_size
+        with input_cols[0]:
+            skip_seconds = st.number_input(
+                "skip_seconds",
+                min_value=0,
+                value=st.session_state.get('skip_seconds', config.frames.get('skip_seconds', 0)),
+                help=tr("Skip the first few seconds"),
+                key="skip_seconds_input"
+            )
+            st.session_state['skip_seconds'] = skip_seconds
+            
+        with input_cols[1]:
+            threshold = st.number_input(
+                "threshold",
+                min_value=0,
+                value=st.session_state.get('threshold', config.frames.get('threshold', 30)),
+                help=tr("Difference threshold"),
+                key="threshold_input"
+            )
+            st.session_state['threshold'] = threshold
+            
+        with input_cols[2]:
+            vision_batch_size = st.number_input(
+                "vision_batch_size",
+                min_value=1,
+                max_value=20,
+                value=st.session_state.get('vision_batch_size', config.frames.get('vision_batch_size', 5)),
+                help=tr("Vision processing batch size"),
+                key="vision_batch_size_input"
+            )
+            st.session_state['vision_batch_size'] = vision_batch_size
 
     # 生成/加载按钮
-    script_path = st.session_state.get('video_clip_json_path', '')
     if script_path == "auto":
         button_name = tr("Generate Video Script")
     elif script_path == "short":
@@ -231,7 +261,10 @@ def render_script_buttons(tr, params):
         if script_path == "auto":
             generate_script_docu(tr, params)
         elif script_path == "short":
-            generate_script_short(tr, params)
+            # 获取自定义片段数量参数
+            custom_clips = st.session_state.get('custom_clips', 5)
+            # 直接将custom_clips作为参数传递，而不是通过params对象
+            generate_script_short(tr, params, custom_clips)
         else:
             load_script(tr, script_path)
 

webui/tools/generate_script_short.py

@@ -11,9 +11,14 @@ from app.config import config
 from webui.tools.base import chekc_video_config
 
 
-def generate_script_short(tr, params):
+def generate_script_short(tr, params, custom_clips=5):
     """
-    生成 纪录片 视频脚本
+    生成短视频脚本
+    
+    Args:
+        tr: 翻译函数
+        params: 视频参数对象
+        custom_clips: 自定义片段数量，默认为5
     """
     progress_bar = st.progress(0)
     status_text = st.empty()
@@ -62,6 +67,7 @@ def generate_script_short(tr, params):
                 base_url=text_base_url,
                 narrato_api_key=narrato_api_key,
                 bert_path="app/models/bert/",
+                custom_clips=custom_clips,
             )
 
             if script is None: