Every clip stops in the review queue
The pipeline ends at a queue, not at a platform. A rendered clip waits there until a person decides. You see the actual finished video — the same file that would post, with its burned-in captions and mixed audio — plus the score the model gave it and the short reasons behind that score, like "complete ending" or "question hook". Approve, reject or send it back. Keyboard shortcuts are there if you review in volume. There is no setting that empties the queue for you, and no path that skips it.
Scripture is forced into review and cannot bypass it
Recited Quran is matched against a 6,236-ayah corpus and rendered as the ayah with its translation. Any clip containing scripture is flagged QUOTE_RISK, and that flag forces human review. It holds even when automation is switched on and the score is high. This is the one rule in the system with no override, because a misattributed or mistimed verse is not the same kind of mistake as a badly cut clip. If a clip carries scripture, you look at it before anyone else does.
It can only post where you have connected an account
Publishing goes to YouTube, TikTok, Instagram and Facebook accounts you have connected yourself. There is no other exit. Approved clips are scheduled into posting windows — four a day, eight on Studio — so you know when things go out. Each destination reports its own state independently. If TikTok refuses a clip that YouTube accepted, the clip is not marked failed and the working post is not undone; you retry the one leg that failed.
The model suggests, it does not decide
A self-hosted Ollama model scores candidate moments and explains why in a few words. Whisper supplies word-level timings, so clips are cut on complete thoughts rather than at fixed intervals. That is the extent of the model's authority. It ranks and it reasons out loud; it never approves, never schedules and never posts. The reasons exist so you can disagree with them quickly. Transcription and scoring run on infrastructure we host, so your lecture audio is not handed to a third-party model provider.
Rejecting a clip costs you nothing
A token is one source minute, charged on the stretch you selected, not the whole video. Clips you reject before export are never charged. Failed renders are never charged. Re-rendering, cutting more clips from the same lecture and editing a clip do not spend tokens again. The point is that saying no should be free, or review quietly turns into a cost you avoid. Basic gives you 40 tokens over a 7-day trial, which is enough to run a real lecture through and reject most of it.
Questions
Can DeenClipped post clips automatically without me seeing them?
No. Every clip lands in the review queue and waits for a human decision. Approval is what makes a clip eligible for scheduling, and scheduling is what makes it eligible to post. There is no auto-publish setting that removes the review step, and clips flagged QUOTE_RISK for containing scripture cannot bypass review under any configuration.
What does QUOTE_RISK mean and why does it force review?
QUOTE_RISK is the flag set on any clip whose transcript contains recited scripture. It forces that clip into human review regardless of its score or your automation settings. Quran recitation is matched against a 6,236-ayah corpus and rendered as the ayah with its translation, and a verse rendered wrongly is a different order of mistake from a badly timed cut. A person checks it.
Where can my approved clips actually be published?
Only to YouTube, TikTok, Instagram and Facebook accounts you have connected yourself. The app has no other publishing route. Approved clips go into posting windows — four a day, eight a day on Studio. Each destination reports its own outcome, so one platform refusing a clip does not mark the clip failed or undo the destinations that accepted it.
How much of my video and audio leaves my control?
You choose a start and end time, and only that stretch is downloaded and processed — the rest of the video is never fetched. Transcription runs on Whisper and moment-scoring runs on a self-hosted Ollama model, both on infrastructure we run, so lecture content is not sent to an outside model provider. Publishing happens only to accounts you have connected.
Basic includes the whole workflow.
Import a source, generate clips, review every one and publish to your own connected channels. Upgrade when you need more.