Breaking the Information Gap: A Local-First AI Workbench for Intensive Podcast Learning.
打破信息差：一款本地优先、零边际成本的 AI 播客精听与知识管理工具。
在 AI 时代，海外有着大量高质量的科技与商业播客（Podcast），它们是获取前沿认知的重要渠道。然而，对于非英语母语的学习者来说，吸收这些内容存在巨大的信息摩擦：
- 听不懂/跟不上：专业词汇密集，语速过快，直接听非常吃力。
- 记不住/无法结构化：听播客通常是线性的，听完即忘，很难像阅读书籍那样进行“划线”或“做笔记”。
- 转录成本高昂：现有的云端转录服务通常按分钟计费。
PodFlow 旨在解决上述痛点。 它是一个英文播客精听工作台。它遵循 本地优先 (Local-First) 原则，利用你本地硬件，将英文长音频转录为可本地储存的字幕，还可以划线、AI查询、写想法，生成跟随音频的笔记卡片。
-
💰 离线转录，零边际成本
- 重资产本地化：最为昂贵、耗时的“语音转文字”环节，集成 WhisperX 模型，支持调用本地显卡 (GPU) 运行，实现50X-70X转录速度。
-
🧠 AI 上下文查询
- 轻量级云端智能：遇到不懂的术语，选中即可查询。利用商业大模型 API（如 DeepSeek, OpenAI 等）的强大推理能力，根据播客上下文提供精准释义。
- 注：此功能需填入您自己的 API Key（Bring Your Own Key）。因查询 token 消耗极低，且按需调用，成本可控。配置方法详见安装文档。
-
📝 沉浸式“双屏”学习体验
- 视听同步：左侧展示精准的时间戳逐字稿，点击任意句子即可跳转音频。
- 笔记联动：右侧为笔记区域。支持对字幕进行划线、AI查询并生成笔记卡片，笔记卡片会自动挂载到对应的时间点上，形成“音频-字幕-笔记”的双向链接。
-
🚧 工程标准与开发状态
- 架构设计：后端基于 FastAPI，前端基于 React + Vite。
- 开发状态：本项目目前处于 Alpha 开发阶段，Bug 在所难免，功能也尚在完善中。
PodFlow.2.mp4
由于本项目在本地进行AI推理，对硬件有一定要求，涉及 PyTorch、CUDA（GPU 加速）环境配置，以及 API Key 的设置，为了保证您的部署体验，我们将详细指南独立整理。
👉 **请阅读根目录下：INSTALLATION-cn（中文安装配置说明）**
High-quality podcasts are a goldmine of knowledge, but for non-native English speakers or deep learners, consuming them is often inefficient:
- The Comprehension Gap: Complex vocabulary and fast pacing make "casual listening" insufficient for learning.
- The Retention Gap: Audio is linear and fleeting. It's hard to highlight, annotate, or structure thoughts while listening.
- The Cost Barrier: Cloud-based transcription services charge by the minute. For avid listeners who consume hours of content weekly, these recurring fees discourage deep engagement.
PodFlow is designed to bridge these gaps. It is a Local-First Workbench for Intensive Learning. By leveraging on-device AI, it transforms audio into a structured, navigable, and interactive knowledge base.
-
💰 Offline Transcription, Zero Marginal Cost
-
Local Heavy Lifting: The expensive and computation-heavy task of "Speech-to-Text" is handled locally by WhisperX using your own GPU.
-
Process unlimited hours of audio without paying per minute.
-
**🧠 AI Context Lookup **
-
Cloud-Powered Intelligence: Highlight any text to trigger explanations. We use lightweight calls to Commercial APIs (e.g., DeepSeek, OpenAI) to provide context-aware definitions.
-
Note: This feature requires your own API Key (Bring Your Own Key). Usage is on-demand and extremely low-cost. Configuration steps are in the Installation Guide.
-
📝 Dual-Pane Interactive UI
-
Synchronized Experience: Transcript on the left, notes on the right. Click any sentence to jump to that exact moment in the audio.
-
Deep Linking: Highlights and notes are anchored to specific timestamps, creating a permanent link between your insights and the source audio.
-
🚧 Engineering & Status
-
Architecture: Built with Python (FastAPI) and React.
-
Development Status: This project is currently in Alpha. While we use Test-Driven Development (TDD) principles to maintain code quality, please expect bugs and rough edges. It is an experimental tool for developers and early adopters, not yet a polished commercial product.
Because PodFlow leverages local hardware acceleration (PyTorch/CUDA) and requires API configuration, the setup process is detailed separately.
👉 **Please read: TODO Installation & Configuration Guide**
This project is licensed under the MIT License.