VidNote AI - Video Transcripts & Notes by shineacz-code

VidNote AI 是一款本地优先的视频字幕与 AI 笔记浏览器扩展，可将 Bilibili 和 YouTube 视频字幕整理为结构清晰、带可跳转时间戳的 Markdown 笔记。

主要功能：
VidNote AI is a local-first browser extension that turns Bilibili and YouTube transcripts into well-structured Markdown notes with timestamps that link back to the original video.

Key features:

• Read the full video transcript in a side panel, choose an available subtitle track, search transcript text, and follow the current playback position.
• Click a timestamp in the transcript or a generated note to seek to the corresponding point in the source video.
• Import local SRT and VTT subtitle files.
• Generate detailed video notes or concise summaries with DeepSeek or another OpenAI-compatible API configured by the user.
• Process the complete transcript in one request, or use chunked parallel generation for long videos with merging, pause, resume, retry, and saved partial progress.
• See the current AI stage, elapsed time, streaming output, and the status of every parallel transcript chunk.
• Save multiple notes for the same video and expand, collapse, copy, delete, or export each note.
• Preview common Markdown elements, including heading navigation, lists, tables, blockquotes, task lists, links, and code blocks.
• Export Markdown compatible with Obsidian Media Extended so timestamp links can reopen the matching position in the source video.
• Use a Simplified Chinese or English interface with light, dark, or system appearance.

VidNote AI does not provide or operate an AI proxy. Users supply their own API endpoint and API key and are responsible for any fees charged by their selected provider. API settings, transcripts, jobs, and notes are stored in the current browser by default. Prompts, the video title, and transcript content are sent directly to the selected AI provider only when the user explicitly tests the connection or starts note generation.

Transcript availability depends on whether the video platform exposes an accessible subtitle track for the current page and account. Output quality, speed, context limits, output limits, and cost depend on the model and provider selected by the user.
- 在侧边栏中读取并展示视频的完整字幕，支持字幕轨选择、关键词搜索和当前播放位置标记。
- 点击字幕或笔记中的时间戳即可跳转到原视频对应位置，方便核对来源和反复学习。
- 支持导入本地 SRT、VTT 字幕文件。
- 使用 DeepSeek 或用户自行配置的 OpenAI-compatible API 生成详尽视频笔记或精简总结。
- 支持整份字幕一次性生成，也支持长视频分段并发处理、自动合并、暂停、恢复、失败重试和部分进度保存。
- 显示 AI 当前阶段、运行时间、并行字幕块状态和流式生成结果，让用户清楚了解处理进度。
- 同一视频可以保存多份笔记，并支持折叠、复制、删除和导出。
- Markdown 预览支持标题导航、列表、表格、引用、任务列表、链接和代码块。
- 可导出兼容 Obsidian Media Extended 的 Markdown；时间戳链接可从 Obsidian 返回原视频位置。
- 提供简体中文和英文界面，以及日间、夜间和跟随系统主题。

VidNote AI 不提供 AI 中转服务。用户需要填写自己的 API 地址和 API Key，并自行承担所选 AI 服务可能产生的费用。API 配置、字幕、任务和笔记默认保存在当前浏览器中。只有在用户主动测试接口或生成笔记时，必要的提示词、视频标题和字幕内容才会直接发送给用户选择的 AI 服务商。

字幕能否读取取决于视频平台是否向当前页面和账号提供可访问的字幕轨。AI 输出质量、速度、上下文长度、输出限制和费用由用户选择的模型及服务商决定。