Releases: nmbrthirteen/podcli
Releases · nmbrthirteen/podcli
v1.0.0
First stable release of podcli — transcribe, clip, and publish short-form video from one place.
License: podcli is now AGPL-3.0 (changed from MIT in this cycle). If you distribute or run a modified version as a network service, the AGPL terms apply.
Highlights
Config profiles & portable setup (#1)
- Split config home (
.podcli/— knowledge, presets, assets, corrections, integrations) from the data dir (data/— cache, transcripts, renders), overridable viaPODCLI_HOME/PODCLI_DATA. - Export / import / activate config bundles to move your whole show setup between machines. Imported bundles no longer carry the exporter's absolute paths.
- Auto-migration of legacy
.podcli/cacheandpresets/into the new layout, idempotent and non-blocking.
Integrations plugin system (#1)
- Pluggable, category-based integration registry exposed over MCP and the web UI, with enable/disable state.
DaVinci Resolve export — beta (#1)
- Export shorts as a Resolve-importable FCPXML timeline; each short becomes an editable compound clip with source on V1 and a ProRes 4444 alpha caption overlay on V2.
- Emits canonical NTSC frame durations and DF/NDF timecode.
⚠️ Beta: the FCPXML math is unit-tested but end-to-end import into Resolve 20.x hasn't been broadly verified yet. Please report issues with your Resolve version and source format.
Reliability
- Non-interactive / CI safe (#3):
podcli processno longer crashes withOSError [Errno 22]when run with piped stdin or no TTY — it renders all suggested clips headlessly instead. - Resume caching: suggestion sessions are reused across runs, keyed on the inputs that affect selection (and skipped when a custom transcript is supplied).
- Repo hygiene: CODEOWNERS for review routing (#2), frame-seam fix for split-screen thumbnails/crops.
Notes
- Requires
ffmpeg(with libass for caption burn-in). - 299 Python + 41 TS tests passing; TypeScript typecheck clean.
Full changelog: v0.1.0...v1.0.0
v0.1.0 - Initial release
Podcli is an open-source AI podcast clipper. Runs on your laptop. One command turns a long-form episode into vertical clips with face tracking and burned-in captions.
What's in it
- One-command pipeline -
./podcli process episode.mp4does the whole thing - AI clip suggestion - Claude or Codex picks moments scored on 4 dimensions
- Face tracking - YuNet detection with smoothed camera, handles split-screen layouts
- 4 caption styles - branded, hormozi, karaoke, subtle (burned in)
- Hardware-accelerated export - VideoToolbox on Mac, NVENC on NVIDIA, VAAPI on Linux
- 3 interfaces - CLI, MCP server for Claude Code, drag-and-drop web app
- Knowledge base - 13 markdown files teach the AI your show's voice
- PodStack workflow - slash commands for titles, descriptions, thumbnails, publish checklists
Install
git clone https://github.com/nmbrthirteen/podcli.git
cd podcli && ./setup.sh
Links
- Website: https://podcli.com
- Quick start: https://github.com/nmbrthirteen/podcli#quick-start
- MCP setup: https://github.com/nmbrthirteen/podcli#mcp-server-claude-integration
Free. MIT licensed.