Make your first clip
Install the prerequisites and configure your API key first. Run commands from the repository root.
See an example first
Open the homepage example to watch a silent generated clip, or start the dashboard and choose Download the example bundle. The example needs no API key and makes no API calls. Its captions and edit are scripted; it demonstrates rendering and exported files, rather than AI selection or editorial acceptance.
Process a recording
.venv/bin/praisonai-clips doctor
.venv/bin/praisonai-clips run 'input/main video.mp4'
The default target is up to five non-overlapping clips, 30–60 seconds each, rendered as 1080 × 1920 MP4. A job can return fewer clips when suitable moments or review evidence are insufficient. Normalizing the source takes CPU time before transcription starts.
For a short initial test, prepare the first four minutes of a local source:
ffmpeg -i '/absolute/path/to/video.mp4' -t 240 \
-c:v libx264 -c:a aac 'input/first-four-minutes.mp4'
.venv/bin/praisonai-clips run 'input/first-four-minutes.mp4' --max-clips 1
Use a new output filename if that file already exists.
Use the dashboard
.venv/bin/streamlit run app.py
Open the local dashboard. On your first visit, add a recording directly in the main screen. Open Use a local file or adjust clip settings for local paths, clip count, and duration. Uploads support up to 2 GB; use a path for larger files. Create my clips becomes available when a source is selected and processing prerequisites are ready.
Review the output
.venv/bin/python scripts/serve_preview.py
Open the preview gallery. Review framing and captions, and listen before publishing. Uncertain results remain in drafts/; accepted clips are in accepted/. The preview server exposes the local data directory on localhost.