Repository navigation
Adding a show
Add a [shows.<key>] table to clipper.toml (see the end of clipper.example.toml):
channels = ["<kick slug>"], [shows.<key>.meta] show_name, and
[shows.<key>.scoring] stream_context / known_figures. For Shorts, measure the show's
screen layouts into [shows.<key>.shorts.layouts.<name>] (panels of the frame, as fractions), one per
scene the show switches between: a Short that crosses a scene cut changes layout there.
A panel wider than its slot is cut down around its focus = [x, y] (the middle by default);
give a webcam panel focus = "face" and each Short's crop is centred on the face
clippa faces found in that box during the scene, so a streamer who sits off-centre or
moves about isn't cut off (the stream needs its faces.json; without it the crop keeps the middle;
the worker renders the Shorts again right after its faces stage, since the fast lane's come first).
Mark every webcam panel cam = true (a focus = "face" panel counts already): the
review checks look for a face there in each Short, and clippa doctor
warns about a show that marks none. The mark doesn't change the picture.
Save a reference screenshot of each one from a recorded stream (clippa shorts <stream> --save-example <layout> --at <seconds>); clippa doctor warns while any is missing.
A layout that frames an existing scene another way (say, only the webcams of a watch scene)
gets pick = false: scene detection never chooses it and it needs no screenshot; pick it by
hand with clippa shorts <stream> --items <vNN> --layout <name>, the Kitsu Layout column or a retake.
Add the slug to [record] channels to record it live. clippa shows lists every show and
its streams; the show's Kitsu production is created on its first publish.
Three [shows.<key>.scoring] settings decide what a Short may be (see [scoring] in clippa/defaults.toml):
third_party_audio (reject, or allow_reactions for reaction clips the show's own
commentary carries, with at most max_source_audio_s of the source), audience_knowledge
(strangers, or fans who know known_figures and their running drama) and, from prompt
v7, gambling (reject, or commentary: gambling as a topic is fine and only a Short that
promotes gambling is rejected). From v7 a post, chat message or statement the show's own
people read aloud is their own audio, not third party, matching the other-voices hold.
clippa models gold grades a model and prompt against
windows you labelled (Model providers).
If the show posts a pre-show on YouTube, add [shows.<key>.preshow] youtube_channel = "UC..." (the channel id: the channel_id in yt-dlp -J <any video url>, or the
channelId in the channel page's source). clippa preshow list should then list its
recent pre-shows, and clippa doctor checks the feed. clippa serve polls it ([preshow]
poll_minutes) and runs new pre-shows on its machine, transcribe included, so serve's
machine needs the GPU extra, Modal, or [transcribe] device = "cpu"; set poll_minutes = 0
on a serve that shouldn't. yt-dlp needs deno for YouTube; uv sync installs it (yt-dlp's
deno extra). Without it YouTube answers downloads with HTTP 403 partway through.
Home · Getting started · clippa doctor · These pages are built from docs/wiki/ in the repo, so edits made here are overwritten: how to edit them
Run it
- Overview
- Getting started
- Roles and services
- Configuration
- Pipeline
- Review
- Reviewing
- War room
- Clippa messenger
- Scheduling and uploads
- Costs
- Checkup
- Audit
- Backups
Install
- Install
- Install checklist
- Kitsu
- Discord app
- YouTube sign-in
- Windows
- macOS
- Transcribe on Modal
- NAS recorder
- NAS hub
- Claude on the NAS
- Phone access
- Model providers
- Adding a show
- YouTube channel
Features
- Pre-shows
- Fetching VODs
- Speakers
- Faces
- Covers
- Cast library
- B-roll
- Clip art
- Weekly issue
- X posts
- TikTok
- News watcher
- Character pieces
- Documentaries
Research
Working on Clippa