Podcli picks the viral moments in your episode, crops them 9:16 with face tracking, burns captions in, and exports upload-ready Shorts. The whole pipeline runs from one binary on your laptop, and you can drive it from Claude Code through an MCP server.
Free tier watermarks exports. Paid tier caps monthly minutes. A weekly show runs out.
Upload is required. Fine for hobby episodes; not fine if anything you're shipping is under NDA or unreleased.
No CLI. No API on the cheap tiers. Nothing an AI coding agent can call.
Pick from the template library. There is no fifth style. There is no editing the template.
Five projects are actively pitched as open-source OpusClip alternatives. They are shaped differently. Pick by the shape that matches what you need.
Full pipeline. Picks moments, crops 9:16 with face/mouth tracking, burns captions via Remotion, exports. Scores against a markdown knowledge base. MCP-native for Claude Code / Codex / Cursor.
The "fuck OpusClip" project. Open source with a hosted upsell. Good shape if you want a Docker-deployable team tool.
Clip-detection primitives and dynamic 9:16 reframing in Python. You build the UI, captions, export queue, ranking around it.
Free, unlimited self-host alternative for AI YouTube Shorts generation. Smaller community.
Whisper + Gemini + MediaPipe / YOLO. Heavier on production polish (B-roll, BGM, thumbnails) than on developer ergonomics.
Five stages in one install. Three ways to call them.
Scripts and batch runs.
Live 9:16 phone preview with optional TikTok wireframe.
19 tools. Plug into Claude Code, Codex, Cursor.
Three things you give up by running it locally.
The questions people search before they pick a clipper.
It depends on the shape you need. Podcli is a full pipeline (CLI + web UI + MCP server) for single creators on a laptop. SupoClip is a Docker-deployable clipper for teams with a hosted upsell. ClipsAI is a Python library for developers composing their own product. ClippedAI and opensource-clipping are smaller scripts.
Yes. Podcli is MIT licensed and exports clean MP4s with no watermark on every run. Files stay on your laptop. There is no minute cap and no paid tier.
Three reasons most people switch: free tier watermarks plus paid-tier minute caps, uploading podcast files to a third party (NDA / unreleased content problem), and the inability to script it from an AI coding agent.
No. Hardware encoders are used when available (VideoToolbox on Mac, NVENC on NVIDIA, VAAPI on Linux), but the CPU fallback works on any modern laptop.
Yes. Podcli ships as an MCP server with 19 tools. Plug it into Claude Code, Claude Desktop, or Codex and say 'clip this episode.' The agent handles transcription, scoring, cropping, captions, and export end to end.
The setup script handles the toolchain. You'll have a clip out the other side in a few minutes.