Podcli turns a long podcast episode into short social clips. It transcribes the recording, scores candidate moments, crops each clip to whoever is speaking, and burns in word-timed captions. Transcription and rendering run on your own machine, and the only network calls are optional: Claude or Codex requests for AI clip scoring, AssemblyAI as an alternative transcription engine, and YouTube publishing.
The same engine has three front ends: a CLI for batch and scripted runs, a browser-based studio with a library, episode workspace, thumbnails, analytics, and an MCP setup page, and an MCP (Model Context Protocol) server with 30 tools so an agent such as Claude Code or Codex can transcribe, score, render, and publish through conversation. PodStack ships alongside it as a set of Claude Code slash commands that take a transcript to a publish-ready package of titles, descriptions, thumbnail briefs, and a checklist.
Beyond clips, Podcli edits full episodes. A multicam edit syncs camera and microphone files by audio, corrects clock drift between devices, and cuts to the active speaker, with split screens for remote shows. The result renders as an MP4 with one WAV per person, or exports as a timeline for Premiere, Final Cut, or DaVinci Resolve. A knowledge base keeps titles and captions in the show's voice and stops the engine from resuggesting moments that were already published.
Features
- Clip formats: 9:16, 16:9, and 1:1 output, with captions sized for each canvas
- Speaker tracking: face tracking follows the speaker, including split-screen layouts
- Tight cuts: multi-segment clips that drop filler, long pauses, and tangents
- Caption styles: branded, hormozi, karaoke, and subtle styles, plus logos, intros, outros, and music from an asset library
- Transcription: Whisper with speaker diarization, AssemblyAI as an alternative, or your own TXT, SRT, or VTT transcript
- Moment scoring: AI scoring against your knowledge base, plus audio energy and laughter detection for highlight reels
- Multicam edit: syncs every camera and mic by audio and cuts to whoever is talking
- Agent control: 30 MCP tools for Claude Code, Codex, Claude Desktop, and other MCP apps
- Publishing: YouTube upload with performance analytics, and FCPXML export for DaVinci Resolve
- Hardware encoding: VideoToolbox, NVENC, and VAAPI, with a CPU fallback and loudness-normalized audio
