AutoClipFirst clips with AutoClip: start with captions or local transcription
Start with 10–30 minutes of footage you own or have permission to use. Complete the input, analysis, rendering and export-kit workflow before processing a whole podcast. AutoClip is an open-source desktop tool with local editing and rendering, and a choice of cloud or local models.
1. Install the official release and check your machine
Download the package for your machine from the website or GitHub Releases. Current desktop packages target Apple Silicon macOS and Windows x64; consult the repository installation guide for operating-system requirements and troubleshooting. Your first run may need a transcription-model download and free disk space. Installing the app is only the first step.
2. Configure and test an analysis model
Choose a provider in Settings, enter your own API key and test the connection. Ollama and LM Studio are local alternatives. The application is free; cloud API usage follows the provider's billing. Disable optional cloud transcription, reference-frame analysis and image generation for an easier first test.
3. Choose input based on available captions
Use a YouTube or Bilibili link, or a local video. Prefer creator-provided captions when available; a local file can use a matching SRT. Without captions, use an installed local transcription model or explicitly configure cloud transcription. Cloud transcription sends audio. Check caption text, timestamps and speaker attribution.
4. Choose one destination and produce
Start with one destination: Douyin, Xiaohongshu or YouTube Shorts. The destination affects framing, captions, covers and post copy. For footage under three minutes, try Shorts or Bilibili first; the current YouTube long-form destination requires a complete segment of at least 180 seconds. Runtime depends on source length, captions, models and hardware.
5. Review complete clips and save the kit
Watch every clip. Keep the question and answer complete, captions clear of faces and important details, the opening understandable and the ending intact. Save the video, cover and post copy, and check that the title reflects the source. Review before publishing; platform uploads happen when you choose a publishing action.
If it stalls, identify the stage
For downloads, check the link and network. For transcription, confirm the model finished downloading and try a matching SRT as a control. For model calls, test the connection and balance. For rendering, record the destination, source details, version and logs. Never share API keys. An installer click, a successful installation and a first successful clip are separate outcomes.