AutoClip moves editing onto your own machine: paste a link, pick platforms, and get videos, covers and post copy sized to each one
Synopsis
AutoClip is an MIT-licensed short-video tool that edits and renders on the local machine: after a link is pasted and platforms chosen, it generates videos, covers and post copy to each platform's aspect ratio and duration, scoring up to 10 qualifying clips per platform and listing the rest as backups. The analysis model is selectable, subtitles can come from the author's subtitles, local Whisper / SenseVoice or a configured cloud transcription, and cloud analysis, transcription and AI image generation are billed by the chosen provider while the software itself is free.
Interpretation
Framing, cutting and rendering of short videos now happen on the local machine, and pasting a link plus choosing platforms yields finished clips, covers and post copy. The earlier path sent material to a cloud service; AutoClip keeps this chain on the user's own computer and sends content only when cloud analysis, cloud transcription or reference-frame image generation is chosen. Three different source materials were tested during development on the same Apple Silicon Mac on 2026-10-01; Jensen used the author's subtitles and TIM used local Whisper base, and the three are not a same-input speed comparison.
Output is adapted per platform: Douyin, Xiaohongshu, TikTok, Reels, YouTube Shorts, Bilibili and YouTube each get clips generated to their aspect ratio and duration requirements. Each platform automatically generates up to 10 qualifying clips by score, with the rest listed as backups to be made by clicking "generate this one"; every clip carries a cover, title, description, topics and a ZIP publishing package. Douyin and Xiaohongshu default to an interview layout, TikTok, Reels and Shorts default to a full-screen podcast layout, portrait output can also choose either layout, and Bilibili and YouTube are landscape, with subtitle and copy language following the platform.
Desktop, CLI and MCP share the same one-click production chain, so agent clients can call the same production entry points. The new MCP entries are start_quick_output and get_quick_output_status, while the older run / export and older clipping tools remain compatible. CLI / MCP requires Python 3.10+ (3.11 recommended) with FFmpeg and FFprobe on PATH, and the local CLI needs no Redis; Intel Mac and Linux have no desktop installer and use Docker or the CLI instead.
Perspective
The demonstrated scope is desktop use on macOS Apple Silicon and Windows 10 / 11 x64 with interview and podcast material, plus Docker / CLI deployment; Intel Mac and Linux currently have no desktop installer. The material addresses users who want editing and rendering done locally and output generated automatically per platform.
The material states that the three tests are not a same-input speed comparison, that the old chain generated 33 clips while the new chain automatically generated 10, and that source materials and output counts differ; costs are estimated from qwen-plus text-model usage at the time and exclude cloud speech recognition, AI image generation and submission services, with actual bills set by the chosen provider.
