Multi-source transcript merging inspired by textual criticism — LLM adjudicates multiple Whisper, YouTube captions & external transcripts for higher quality. Includes speaker diarization and summarization.

构建你绝对掌控的私人播客金库，告别平台的封闭限制。 一键批量下载、AI智能提取，让每一期喜爱的小宇宙播客真正沉淀为你自己的知识资产。

Agent skill for multi-speaker meeting & podcast transcription — FunASR (Paraformer/SenseVoice/Whisper) + MiMo-V2.5-ASR + CAM++ speaker diarization + LLM cleanup. 99 languages, GPU & CPU. Works with Claude Code and any SKILL.md-compatible agent.

Xiaoyuzhou/小宇宙 + podcast/YouTube transcription with speaker diarization. 小宇宙/播客/YouTube 转写：声纹区分 + 连续段落稿。

Production pipeline around OpenAI gpt-4o-transcribe-diarize for long-form 2-speaker interviews. Cross-chunk speaker consistency · diarization hallucination fix · async GPT-5.5 domain-term correction. WER 6.05% / DER 4.28% on 2h26m benchmark. Beats raw OpenAI API by +11.5 Q.

Cross-platform CLI to download & transcribe podcasts locally — Apple Podcasts, Xiaoyuzhou, RSS feeds with built-in Whisper speech-to-text (Metal/CUDA/CPU)

AI-powered content platform that turns RSS feeds, podcasts, YouTube videos, articles, GitHub repos, and images into summaries, generated articles, and platform-ready social posts. Transcribes with Whisper, generates with OpenAI or local Ollama, and schedules/publishes to LinkedIn, Threads, Facebook, and X/Twitter from one self-hosted Flask UI.