AI video clipping · for podcasts, interviews & sermons

One long video in.
A week of clips out.

Paste a YouTube link or drop in a file. Decaris AI finds the moments that stand on their own, cuts them on sentence boundaries, reframes them vertical and burns in word-by-word captions — ready for TikTok, Reels and Shorts.

1080×1920 · H.264 −14 LUFS audio karaoke captions title · hook · hashtags
Vertical clip of an interview with the caption 'don't you believe in'
Vertical clip of a podcast guest with the caption 'personal portfolio, how'
Vertical clip of a podcast guest with the caption 'basically democratized'
Real output · blurred-background reframe

How it works

It reads the whole video, so you don't have to scrub it.

Every word is transcribed with its own timestamp. Then Claude reads the transcript window by window and picks the moments with a hook, a turn and a payoff — not just the loud bits.

SOURCE · finance podcast, 1 h+ 11 clips · spaced so no two cover the same moment
Top 3 by score Other selected clips minutes into the source
1 · Bring the video

Link or upload

Paste a YouTube URL or drag in a file from your camera, Zoom or podcast recorder.

2 · Transcribe

Word-level timestamps

Whisper transcribes every word with its exact timing, in the language spoken. Transcripts are cached, so re-runs are instant.

3 · Pick the moments

Claude chooses

Each candidate is scored 0–100 on how well it plays with no context — and given a title, a hook and hashtags.

4 · Render

One pass, done

Cut, reframe, caption and loudness-normalise in a single render — about 5 seconds per clip.


The app

One screen. One button. Create clips.

No timeline editor, no keyframes, no export dialogs. The defaults are already tuned for vertical short-form — most people never open the options.

The Decaris AI app: a video URL field, a file upload area, clip options on the left and a preview and clip list on the right
Runs on your own Mac — your footage never leaves your machine to be edited. Snapshot of the live app

Features

Built by people who actually post clips.

Moment selection

Claude picks what's worth watching

It reads the entire transcript in windows, so a great answer at minute 57 is found as easily as one at minute 3. Every clip comes back with a reason and a score.

Captions

Word-by-word karaoke captions

White text with the spoken word lit in yellow, burned into the video and placed above the TikTok UI — never hidden behind the username or buttons.

Reframing

Nobody gets cropped out

Blurred-background framing keeps both people on a two-person set in shot. Prefer full-bleed? Switch to centre crop.

Clean cuts

Never mid-sentence

Both edges snap to sentence boundaries and cuts are frame-accurate, so captions stay in sync from the first word to the last.

Audio

Mastered to −14 LUFS

The loudness streaming platforms target. No clip is suddenly quiet or blown out next to the one before it.

Post-ready

Titles, hooks & hashtags written

Every clip ships with copy chosen for that specific moment, saved in a clips.json beside the videos — copy, paste, post.

In control when you want it

Every setting is a real decision, not a knob.

Aspect ratio

9:16 for TikTok, Reels and Shorts. 1:1 for feed posts. 16:9 to extract the moment as-is.

Clip length

15–30s for punchlines, 40–60s for a full thought, up to 120s for arguments that need the room.

Number of clips

1 to 20. Spacing rules stop it handing you six takes of the same moment.

Language

Auto-detects the spoken language, or set it yourself.

Decaris AI options: aspect ratio, reframing, captions, Claude moment selection, download quality and the Create clips button

What you get

A folder of clips you can post today.

Frames below are straight from Decaris AI renders — no retouching. Each clip lands as its own MP4 with its hook, timecode and score recorded next to it.

Swipe for more clips →

"file": "Clip_01.mp4",
"hook": "We're getting too big for this small pond",
"score": 58,
"start": 3461.59,  "end": 3519.91,
"title": …,  "hashtags": [ … ]

Who it's for

If you record long, you should be posting short.

PodcastersTurn every episode into a week of discovery clips.
ChurchesShare the sermon's strongest minute with people who missed Sunday.
EducatorsCut a lecture into bite-size explanations students actually finish.
AgenciesDeliver short-form for every client without an editor per account.
InterviewersSurface the quote everyone will share — with the context intact.

Questions

Straight answers.

How long does it take?

Transcription is almost all of the wait: roughly 3–4 minutes for a 13-minute video on the default setting. Rendering is about 5 seconds per clip, three at a time. A 90-minute podcast is a coffee break, not a lunch break.

Do I need to edit anything afterwards?

No. Clips come out captioned, vertical and loudness-matched. Download the ones you like and upload them straight to TikTok, Instagram or YouTube.

What videos work best?

Anything where people talk: podcasts, interviews, debates, sermons, lectures, panels. You can paste a YouTube link or upload your own file.

Does it work in languages other than English?

Yes. The spoken language is detected automatically, or you can set it yourself.

Where does my video go?

Decaris AI runs on your own computer. Your footage is downloaded, transcribed and rendered locally; only the transcript text is sent to Claude to choose the moments.

Stop scrubbing timelines

Your next ten clips are already in last week's video.

Decaris AI is designed and built by MkstanCreative.

mkstancreative.online ↗
Chat on WhatsApp