Use case · Long to short
Say what the short
must say.
An hour of recording usually holds less than a minute that matters to the audience you have in mind. Long video to short video AI, as Anjin Media builds it, starts with you describing that minute - the point it must make, who it is for, how long it runs - and returns a finished cutdown composed from the moments that make the case, with source timecodes on every segment.
The same workflow covers a podcast episode, a webinar recording or a keynote master: anything long, spoken and yours.
One video or an entire archive.
Long video to short product demonstration
The problem
The highlight lottery.
Feed a recording to a conventional clipper and it returns candidates: a page of scored moments, ranked by how much each one resembles a clip that once performed. None of them knows what you are trying to say. The work of making a point - choosing an opening, cutting the aside, landing the ending - comes back to you, once per candidate.
A short worth publishing is an editorial decision, and the instrument should accept editorial instruction. What you want to hold at the end is an intentional cutdown, not a lottery ticket - an edit that had a job before it existed.
How composition differs from clipping, layer by layer - the category page walks all five →
What you make
Three shorts, three jobs.
The podcast pull-quote
One exchange from an episode, cut to forty-five seconds for Shorts and Reels. Turning a podcast into short clips starts by naming the exchange; finding where it lives in eighty minutes of transcript is the platform's job.
0:45 · 9:16 · STACKED CROPS
The LinkedIn argument
Ninety seconds from a webinar or conference talk that makes one case start to finish - set up, evidence, landing - captioned for feeds that play silent.
1:30 · 1:1 OR 16:9 · SRT SHIPPED
The YouTube-to-Shorts cut
A published long-form video, re-briefed as verticals: upload the master file and describe each Short as its own edit. One plan can render widescreen, square and vertical together.
UP TO 3 ASPECTS PER EDIT
Starting from a webinar and needing a whole campaign, promo to summary? That workflow has its own page →
A worked example
The brief does the scrubbing.
This is the whole interface: a paragraph about the short and a handful of settings. A brief can run to 2,000 characters - room to say who the clip is for, what it must include and what it must leave out.
What comes back first is a cut plan: the chosen segments in order, each with its transcript text and source timecodes, before anything renders. Read it, strike a segment, ask in plain language for a different opening. Source minutes are counted once, as you bring footage in. Briefing, plans, revisions and renders use none.
Planning is asynchronous - the request is accepted at once and the plan follows when it is composed, so briefs queue while you work. The render arrives with SRT captions, an EDL in JSON and a QC report attached.
“A 45-second vertical for Shorts: the exchange where Maya explains why they shut down their first product. Open on the host’s question, keep her answer intact, end on the line about charging more sooner. No cold open, no montage.”
Where it fits
Right tool, stated conditions.
CHOOSE IT WHEN
- You know what the short must say - or want a readable plan to decide it from.
- The source is spoken: interviews, podcasts, webinars, talks, lectures.
- You need exact lengths and several formats - a target from 10 to 600 seconds, up to three aspect ratios off one plan.
- Provenance matters: every segment must be checkable against the source.
LOOK ELSEWHERE WHEN
- The footage is wordless. Retrieval runs on the transcript, so action reels and b-roll montages are a different tool's job.
- You want a daily stream of auto-picked clips with no direction attached.
- A single file runs past four hours - split the recording into sessions first.
THE EDGES, PLAINLY
Three limits worth knowing before you start. A source file caps at 16 GB and four hours, so a full-day recording loads as separate sessions. Plans take a moment: the brief is acknowledged immediately and the plan follows once composed. And vertical framing today is bars - the full frame, letterboxed - or stack, two crops sharing the frame; a crop that follows the active speaker is on the roadmap and is not in the platform yet.
The objection
Asked straight, answered straight.
- Why not run a clipper and pick the best of the batch?
- When one lucky moment is all the short needs, a clipper may well hand it to you. But a short that must say something specific - set up, evidence, landing - is rarely one continuous stretch of tape. Composition selects across the whole runtime and orders the pieces to your brief, and the plan shows every chosen segment with its source timecodes before you render. The judgement stays yours; the scrubbing goes.
Questions
- How do I turn a podcast into short clips with AI?
- Upload the episode - a source file can run to 16 GB and four hours - and write a brief that says what each clip must say, with a target length between 10 and 600 seconds. The plan comes back as ordered segments with transcript text and source timecodes; when you render, the clip ships with SRT captions, an EDL in JSON and a QC report.
- Can AI turn a YouTube video into Shorts?
- Yes, working from your own master: upload the file you published (mp4, mov or mkv) and brief each Short as its own edit. Vertical renders at 1080×1920 with the full frame letterboxed or two stacked crops, and the same plan can render 16:9 and 1:1 versions alongside.
- How long can a video be for AI to shorten it?
- Up to four hours and 16 GB per file, with up to four cameras in a source group. A longer recording works split into sessions - and one brief can draw on up to four of them at once.