Repository navigation
Scheduling and uploads
Approved items get release slots (clippa schedule, written to schedule.json) and go to
YouTube as private videos with a publish time (clippa upload), or stay private for you to
publish by hand with [youtube] manual_publish. Everything uploads to one YouTube channel;
bringing a new one up is YouTube channel.
Note
Uploads are automatic: don't run clippa upload or clippa review publish by hand
unless you mean to.
clippa serve replans and uploads every [youtube] auto_upload_minutes (the default 0 means
only by hand, and clippa doctor says so on a serve machine), at most [youtube] max_per_24h in any 24 hours, so hourly runs stay under the
channel's daily upload limit. Pre-shows and stories set to "asap" go out on the next run,
outside the release slots.
clippa status is the one screen for all of this: the services' heartbeats and the recorder,
the worker's streams, review counts, the next releases and the month's spend (--days N looks
further ahead; --sync reads Kitsu first).
To stay current rather than build a backlog, [schedule] stream_budget caps each stream's
releases per kind (scaled by its length; the best Shorts by score, segments by their
segment score with chat heat breaking ties, or by heat where a
stream has no score, and compilations by their best segment) and stale_days lets go of whatever isn't out that many days after it could
first go. Items let go stay approved in Kitsu (war room filter "Not going out"); pin one with
"Publish at" to send it anyway. A stream_budget kind at 0 goes only when pinned. A big night
can jump the queue with [schedule] priority = true in its stream config
(Configuration).
In the war room the schedule shows the open slots of the next two weeks; drag an item onto a slot to pin it.
The budget is only as good as what it ranks by. clippa analytics --lift checks each ranker
against how the uploads did in their first 7 days (no model call; it reads analytics.json, so
run clippa analytics first if it's stale):
clippa analytics --lift # every show and kind
clippa analytics --lift --show kino --kind segment
clippa analytics --lift --jsonThe rankers are the Score column (Shorts and clips), the chat heat segments and compilations
are ranked by, the topic's track record (the median 7-day views of the show's earlier uploads
on the same topic, counting only those whose 7 days were over when the stream began), and the
segment score: model from segments.json, bt and bt_tx from the
backtest (summaries, and with the transcript).
For each, per show and kind: how many uploads have a value, Spearman's rank correlation with
7-day views (+1 orders them exactly as views do, 0 is no relation, nothing under 5 values),
and, inside each stream with 4 or more, the views per upload of the uploads ranked above the
middle against those below it (uploads tied across the middle count in neither), with the
number of streams where the top half won. Median views per score follow the table. Each show
stands alone: Kino's uploads draw far more than Channel 404's, so a row pooling them would
mostly measure the show.
Only what went out has views, so a ranker is checked on what it let through, not on what the
budget dropped. clippa analytics --report ends with the same table, over every upload so far.
Upload the file in Studio and clippa upload --attach <item> <link> fills in the rest (title,
description, tags, thumbnail, captions, playlist, Kitsu link). clippa upload --next 5 lists
what goes out next with each file.
Descriptions link to the longer video each one comes from (a clip to its segment at the clip's
offset, a segment to its compilation) once that video is public; every upload run adds the
link to videos already up when their longer video goes public ([youtube] source_links).
A Short's "related video" (the link under it to its segment) can only be set in Studio:
clippa upload --related lists the uploaded Shorts whose segment is out, with the Studio link
and the segment id to search for, and clippa upload --related <keys or links> records the
ones you set. The daily digest lists them too.
A new title for an item that's already on YouTube renames the video there too, wherever it's
made: the War room's item window or find and replace, the Discord bot's details
form, the Clippa messenger, or
clippa review retitle <item> <new title>. The title gets the show name and bleeps as an
upload's does, the rest of the video's details stay as YouTube has them, and Kitsu's title
column changes after YouTube's, so a rename that fails changes neither. Undo renames it back.
A retake's planned title never goes to a video that's up: only a person's edit does.
51 quota units each (1 when YouTube has that title already), from the spare only: past it the
edit is refused with a line saying so. A video deleted in Studio gets the Kitsu change alone.
Its description, hashtags, tags and playlist stay Kitsu-only once it's up.
Every YouTube API call is counted per Pacific day in logs/youtube_quota.json; the link
updates and retries of captions that failed to go up (clippa upload --captions) only spend
what is left after room for the day's remaining uploads ([youtube] daily_quota,
upload_units). Details in YouTube channel, "API quota".
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