timerring · timerring · Dec 30, 2024 · Dec 30, 2024 · Dec 30, 2024 · Dec 30, 2024
diff --git a/README.md b/README.md
@@ -4,12 +4,18 @@
     <img src="assets/headerLight.svg" alt="BILIVE" />
   </picture>
 
-*7 x 24 小时无人监守录制、渲染弹幕、识别字幕、自动上传，启动项目，人人都是录播员。*
+*7 x 24 小时无人监守录制、渲染弹幕、识别字幕、自动切片、自动上传，启动项目，人人都是录播员。*
 
 [:page_facing_up: Documentation](https://timerring.github.io/bilive/) |
 [:gear: Installation](#quick-start) |
 [:thinking: Reporting Issues](https://github.com/timerring/bilive/issues/new/choose)
 
+支持模型
+
+  <img src="assets/openai.svg" alt="OpenAI whisper" width="60" height="60" />
+  <img src="assets/zhipu-color.svg" alt="Zhipu GLM-4V-PLUS" width="60" height="60" />
+  <img src="assets/gemini-brand-color.svg" alt="Google Gemini 1.5 Pro" width="60" height="60" />
+
 </div>
 
 ##  1. Introduction
@@ -29,6 +35,7 @@
 - **自动渲染弹幕**：自动转换xml为ass弹幕文件并且渲染到视频中形成**有弹幕版视频**并自动上传。
 - **硬件要求极低**：无需GPU，只需最基础的单核CPU搭配最低的运存即可完成录制，弹幕渲染，上传等等全部过程，无最低配置要求，10年前的电脑或服务器依然可以使用！
 - **( :tada: NEW)自动渲染字幕**(如需使用本功能，则需保证有 Nvidia 显卡)：采用 OpenAI 的开源模型 [`whisper`](https://github.com/openai/whisper)，自动识别视频内语音并转换为字幕渲染至视频中。
+- **( :tada: NEW)自动切片上传**：根据弹幕密度计算寻找高能片段并切片，结合多模态视频理解大模型 [`GLM-4V-PLUS`](https://bigmodel.cn/dev/api/normal-model/glm-4) 自动生成有意思的切片标题及内容，并且自动上传。
 
 项目架构流程如下：
 
@@ -46,8 +53,13 @@ graph TD
         ifDanmaku -->|有弹幕| DanmakuFactory[DanmakuFactory]
         ifDanmaku -->|无弹幕| ffmpeg1[ffmpeg]
         DanmakuFactory[DanmakuFactory] --根据分辨率转换弹幕--> ffmpeg1[ffmpeg]
+        ffmpeg1[ffmpeg] --渲染弹幕及字幕 --> Video[视频文件]
+        Video[视频文件] --计算弹幕密度并切片--> GLM[多模态视频理解模型]
+        GLM[多模态视频理解模型] --生成切片信息--> slice[视频切片]
         end
-        ffmpeg1[ffmpeg] --渲染弹幕及字幕 --> uploadQueue[(上传队列)]
+
+        slice[视频切片] --> uploadQueue[(上传队列)]
+        Video[视频文件] --> uploadQueue[(上传队列)]
 
         User((用户))--upload-->startUpload(启动视频上传进程)
         startUpload(启动视频上传进程) <--扫描队列并上传视频--> uploadQueue[(上传队列)]
@@ -110,15 +122,21 @@ pip install -r requirements.txt
 ./setPath.sh && source ~/.bashrc
 ```
 
-#### 3. 配置 whisper 模型
+#### 3. 配置 whisper 模型及 GLM-4V-PLUS 模型
 
+##### 3.1 whisper 模型
 项目默认采用 [`small`](https://openaipublic.azureedge.net/main/whisper/models/9ecf779972d90ba49c06d968637d720dd632c55bbf19d441fb42bf17a411e794/small.pt) 模型，请点击下载所需文件，并放置在 `src/subtitle/models` 文件夹中。
 
 > [!TIP]
 > 使用该参数模型至少需要保证有显存大于 2.7GB 的 GPU，否则请使用其他参数量的模型。
 > + 更多模型请参考 [whisper 参数模型](https://timerring.github.io/bilive/models.html) 部分。
 > + 更换模型方法请参考 [更换模型方法](https://timerring.github.io/bilive/models.html#更换模型方法) 部分。
 
+##### 3.2 GLM-4V-PLUS 模型
+
+> 此功能默认关闭，如果需要打开请将 `src/config.py` 文件中的 `AUTO_SLICE` 参数设置为 `True`
+
+在配置文件 `src/config.py` 中，`SLICE_DURATION` 以秒为单位设置切片时长（不建议超过 1 分钟），在项目的自动切片功能需要使用到智谱的 [`GLM-4V-PLUS`](https://bigmodel.cn/dev/api/normal-model/glm-4) 模型，请自行[注册账号](https://www.bigmodel.cn/invite?icode=shBtZUfNE6FfdMH1R6NybGczbXFgPRGIalpycrEwJ28%3D)并申请 API Key，填写到 `src/config.py` 文件中对应的 `Your_API_KEY` 中。
 
 #### 4. biliup-rs 登录
 
@@ -176,7 +194,7 @@ logs # 日志文件夹
 ```
 
 ### Installation(无 GPU 版本)
-无 GPU 版本过程基本同上，可以跳过步骤 3，需要注意在执行步骤 5 **之前**完成以下设置将确保完全用 CPU 渲染视频弹幕。
+无 GPU 版本过程基本同上，可以跳过步骤 3 配置 whisper 的部分，需要注意在执行步骤 5 **之前**完成以下设置将确保完全用 CPU 渲染视频弹幕。
 
 1. 请将 `src/config.py` 文件中的 `GPU_EXIST` 参数设置为 `False`。（若不置为 `False` 且则会使用 CPU 推理，不推荐，可自行根据硬件条件进行尝试。）
 2. 将 `MODEL_TYPE` 调整为 `merge` 或者 `append`。

diff --git a/assets/gemini-brand-color.svg b/assets/gemini-brand-color.svg
diff --git a/assets/openai.svg b/assets/openai.svg
diff --git a/assets/zhipu-color.svg b/assets/zhipu-color.svg
diff --git a/src/autoslice/__init__.py b/src/autoslice/__init__.py
@@ -0,0 +1,5 @@
+# Copyright (c) 2024 bilive.
+
+import sys
+import os
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
diff --git a/src/autoslice/calculate_density.py b/src/autoslice/calculate_density.py
@@ -0,0 +1,51 @@
+# Copyright (c) 2024 bilive.
+
+import re
+from collections import defaultdict
+from src.config import SLICE_DURATION
+
+def parse_time(time_str):
+    """Convert ASS time format to seconds with milliseconds."""
+    h, m, s = time_str.split(':')
+    s, ms = s.split('.')
+    return int(h) * 3600 + int(m) * 60 + int(s) + int(ms) / 100
+
+def format_time(seconds):
+    """Format seconds to hh:mm:ss.xx."""
+    h = int(seconds // 3600)
+    m = int((seconds % 3600) // 60)
+    s = int(seconds % 60)
+    ms = int((seconds - int(seconds)) * 100)
+    return f"{h:02}:{m:02}:{s:02}.{ms:02}"
+
+def extract_dialogues(file_path):
+    """Extract dialogue start times from the ASS file."""
+    dialogues = []
+    with open(file_path, 'r', encoding='utf-8') as file:
+        for line in file:
+            if line.startswith('Dialogue:'):
+                parts = line.split(',')
+                start_time = parse_time(parts[1].strip())
+                dialogues.append(start_time)
+    return dialogues
+
+def calculate_density(dialogues, window_size=SLICE_DURATION):
+    """Calculate the maximum density of dialogues in a given window size."""
+    time_counts = defaultdict(int)
+    for time in dialogues:
+        time_counts[time] += 1
+
+    max_density = 0
+    max_start_time = 0
+
+    # Use a sliding window to calculate density
+    sorted_times = sorted(time_counts.keys())
+    for i in range(len(sorted_times)):
+        start_time = sorted_times[i]
+        end_time = start_time + window_size
+        current_density = sum(count for time, count in time_counts.items() if start_time <= time < end_time)
+        if current_density > max_density:
+            max_density = current_density
+            max_start_time = start_time
+
+    return max_start_time, max_density
diff --git a/src/autoslice/slice_video.py b/src/autoslice/slice_video.py
@@ -0,0 +1,61 @@
+# Copyright (c) 2024 bilive.
+
+import subprocess
+from src.autoslice.calculate_density import extract_dialogues, calculate_density, format_time
+from src.config import Your_API_KEY, SLICE_DURATION
+import base64
+from zhipuai import ZhipuAI
+
+def zhipu_glm_4v_plus_generate_title(video_path, artist):
+    with open(video_path, 'rb') as video_file:
+        video_base = base64.b64encode(video_file.read()).decode('utf-8')
+
+    client = ZhipuAI(api_key=Your_API_KEY)
+    response = client.chat.completions.create(
+        model="glm-4v-plus",
+        messages=[
+        {
+            "role": "user",
+            "content": [
+            {
+                "type": "video_url",
+                "video_url": {
+                    "url" : video_base
+                }
+            },
+            {
+                "type": "text",
+                "text": f"视频是{artist}的直播的切片，请根据该视频中的内容及弹幕信息，为这段视频起一个调皮并且吸引眼球的标题，注意标题中如果有“主播”请替换成{artist}。"
+            }
+            ]
+        }
+        ]
+    )
+    return response.choices[0].message.content.replace("《", "").replace("》", "")
+
+# https://stackoverflow.com/questions/64849478/cant-insert-stream-metadata-into-mp4
+def inject_metadata(video_path, generate_title, output_path):
+    """Slice the video using ffmpeg."""
+    command = [
+        'ffmpeg',
+        '-i', video_path,
+        '-metadata:g', f'generate={generate_title}',
+        '-c:v', 'copy',
+        '-c:a', 'copy',
+        output_path
+    ]
+    subprocess.run(command)
+
+def slice_video(video_path, start_time, output_path, duration=f'00:00:{SLICE_DURATION}'):
+    """Slice the video using ffmpeg."""
+    command = [
+        'ffmpeg',
+        '-ss', format_time(start_time),
+        '-i', video_path,
+        '-t', duration,
+        '-map_metadata', '-1',
+        '-c:v', 'copy',
+        '-c:a', 'copy',
+        output_path
+    ]
+    subprocess.run(command)
diff --git a/src/burn/only_render.py b/src/burn/only_render.py
@@ -3,10 +3,13 @@
 import argparse
 import os
 import subprocess
-from src.config import GPU_EXIST, SRC_DIR, MODEL_TYPE
+from src.config import GPU_EXIST, SRC_DIR, MODEL_TYPE, AUTO_SLICE, SLICE_DURATION
 from src.burn.generate_danmakus import get_resolution, process_danmakus
 from src.burn.generate_subtitles import generate_subtitles
 from src.burn.render_video import render_video
+from src.autoslice.slice_video import slice_video, inject_metadata, zhipu_glm_4v_plus_generate_title
+from src.autoslice.calculate_density import extract_dialogues, calculate_density, format_time
+from src.upload.extract_video_info import get_video_info
 import queue
 import threading
 import time
@@ -52,7 +55,20 @@ def render_video_only(video_path):
     render_video(original_video_path, format_video_path, subtitle_font_size, subtitle_margin_v)
     print("complete danamku burning and wait for uploading!", flush=True)
 
-    # # Delete relative files
+    if AUTO_SLICE:
+        title, artist, date = get_video_info(format_video_path)
+        slice_video_path = format_video_path[:-4] + '_slice.mp4'
+        dialogues = extract_dialogues(ass_path)
+        max_start_time, max_density = calculate_density(dialogues)
+        formatted_time = format_time(max_start_time)
+        print(f"The 30-second window with the highest density starts at {formatted_time} seconds with {max_density} danmakus.", flush=True)
+        slice_video(format_video_path, max_start_time, slice_video_path)
+        glm_title = zhipu_glm_4v_plus_generate_title(slice_video_path, artist)
+        slice_video_flv_path = slice_video_path[:-4] + '.flv'
+        inject_metadata(slice_video_path, glm_title, slice_video_flv_path)
+        os.remove(slice_video_path)
+
+    # Delete relative files
     for remove_path in [original_video_path, xml_path, ass_path, srt_path, jsonl_path]:
         if os.path.exists(remove_path):
             os.remove(remove_path)
@@ -63,6 +79,9 @@ def render_video_only(video_path):
 
     with open(f"{SRC_DIR}/upload/uploadVideoQueue.txt", "a") as file:
         file.write(f"{format_video_path}\n")
+        if AUTO_SLICE:
+            print("complete slice video and wait for uploading!", flush=True)
+            file.write(f"{slice_video_flv_path}\n")
 
 class VideoRenderQueue:
     def __init__(self):

diff --git a/src/config.py b/src/config.py
@@ -10,7 +10,10 @@
 # Can be pipeline, append, merge
 MODEL_TYPE = "pipeline"
 Inference_Model = "small"
-
+AUTO_SLICE = False
+SLICE_DURATION = 30
+# Apply for your own GLM-4v-Plus API key at https://www.bigmodel.cn/invite?icode=shBtZUfNE6FfdMH1R6NybGczbXFgPRGIalpycrEwJ28%3D
+Your_API_KEY = ""
 # ============================ Basic configuration ============================
 SRC_DIR = str(Path(os.path.abspath(__file__)).parent)
 BILIVE_DIR = str(Path(SRC_DIR).parent)