You pick the source, and the exact minutes
Paste a YouTube link or upload an MP4 or MOV. Then set a start and an end time; the shortest range you can select is 30 seconds. Only that stretch is downloaded and processed, so taking three minutes out of a ninety-minute lecture costs three minutes of work rather than ninety. Billing follows the same line. One token is one source minute, and you are charged for the range you picked, not for the length of the video you picked it from.
Transcription first, then a search for complete moments
Whisper transcribes the selected audio with word-level timings. You can pin the language to English, Arabic or Urdu, or leave it on auto-detect, which is multilingual and switches per segment, so an English talk containing Arabic recitation is handled rather than flattened into one language. A self-hosted Ollama model then reads the transcript and scores candidate moments, returning short reasons: complete ending, question hook, stands alone. Clips are cut on those complete moments, not at fixed intervals, so a point finishes before the clip does.
Framing, burnt-in captions and a nasheed underneath
Face detection crops the 16:9 source to vertical 9:16 and keeps the speaker in frame. Captions are burnt into the video with libass, timed to the words: word-by-word, karaoke, phrase or stacked lines, depending on the template. There are five templates, being Bold Stack, Clean Line, Headline, Mono Minimal and Quran Recitation. A nasheed is mixed in and ducked under the speech; music is on unless it is explicitly switched off for that job. What comes out is a finished vertical file, not a preview of one.
When the words are Arabic, or scripture
Recited Quran is matched against a 6,236-ayah corpus and rendered as the ayah with its translation, set in Amiri. Arabic speech that is not scripture is captioned in Arabic with an English line beneath it, taken from a second translate pass. The Quran Recitation template captions scripture and nothing else, so a half-heard aside never sits under a verse in the lecture face. Any clip containing scripture is flagged QUOTE_RISK and forced into human review. That gate never bypasses, whatever else is set to run on its own.
Nothing leaves until you approve it
Every clip lands in a review queue and waits there. You watch the rendered clip itself, the same file that would post, with its score and the reasons the model gave for choosing that moment. Approve it, reject it, or send it back; there are keyboard shortcuts for working through a stack. Approved clips fill posting windows, four a day and eight on Studio, and go to the YouTube, TikTok, Instagram and Facebook accounts you connected. Each destination reports its own state, so one platform refusing does not mark the clip failed.
Questions
Does DeenClipped publish clips automatically?
No. Every clip generated lands in a review queue and stays there until a person decides on it. There is no setting that skips that step. Clips containing scripture are additionally flagged QUOTE_RISK and forced through the same review, which is the one gate that never bypasses. Publishing only reaches accounts you connected yourself, and only after approval.
What does one DeenClipped token cover?
One token is one source minute, counted against the stretch you selected rather than the length of the original video. Basic is free: 40 tokens across a seven-day trial. Pro and Studio are paid, billed weekly, monthly or yearly. Re-rendering a clip, cutting more clips from the same lecture and editing all cost nothing extra, and a failed render or a clip you reject before export is never charged.
How does DeenClipped handle Quran recitation?
Recited Quran is matched against a 6,236-ayah corpus and rendered as the ayah with its translation, set in Amiri rather than the lecture caption face. The Quran Recitation template captions scripture and nothing else, so a half-heard aside never appears under a verse. Any clip containing scripture carries a QUOTE_RISK flag that forces a human decision before it can go anywhere.
What can DeenClipped not do yet?
The clip editor trims a clip, cuts sections out of the middle, edits caption words and repositions captions; it has no frame-by-frame tools, overlays or extra media. There are no view counts, watch time or audience figures anywhere in the app, because no connected platform sends them. Posting to two accounts on the same platform is not built yet either. There is no mobile app.
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.