A newsroom clip has one obligation: to show what was said, as it was said, from the start of the sentence to its end. shortshort selects passages that open and close a thought, keeps the original sound and the picture cropped to 9:16 around the face on screen, and lays captions timed on the audio over them. The tool is the shortshort studio itself, not a newsroom edition of it, and it puts nothing on the clip that was not in the recording.
30 free credits, no cardUp to 2 GB per videoMP4 · 9:16 or 4:5
shortshort cuts a press conference or interview into vertical clips chosen to start and end on a complete sentence, cropped to 9:16 around the face on screen and captioned word by word in the language spoken, with nothing added to the picture or the sound. Up to 10 clips per video on Creator, 20 on the Studio plan; no API, and scheduled publishing goes to YouTube only, with TikTok in early access.
Recommended plan: Creator. Creator for a desk under 26 hours 40 minutes of footage a month: 1,600 credits, 10 clips per video, top-ups for a busy month, 29 EUR a month. Past that volume, the Studio plan: 6,000 credits for 100 hours, 20 clips per video and the JSON edit plan of every saved clip, 79 EUR a month, both plus VAT.
Facts checked against the product on . Prices before VAT.
Four facts, each with its number.
The whole recording is transcribed word by word and passages are chosen for being complete and quotable. To finish the sentence, a clip may run 25% shorter or longer than the length requested: a 30-second target can come out anywhere between 22.5 and 37.5 seconds rather than stop halfway through a claim. The AI can join up to 8 excerpts of the recording in one clip; the studio lists each one with its timecodes so the desk can check every cut.
No B-roll is inserted, no music laid under the voice, no title written on the picture by the AI, no dubbing, no translation. The output is the source cropped to 9:16 with captions in the language spoken, so what a viewer hears and reads is what the person said. Zooms exist, bounded between x1.0 and x1.6, and are set automatically.
Each selected passage is transcribed a second time to time every word on the audio; where that pass cannot align, the first-pass timing is kept and marked as estimated so a subeditor knows which lines to check. An SRT already timed to the source can be imported instead. Minimal, the quiet caption style with no effects, is the one that reads like a subtitle.
A 40-minute press conference costs 40 credits whether the desk takes a single quote from it or twenty on the Studio plan. Creator, at 29 EUR a month, holds 1,600 credits, or 26 hours 40 minutes of footage, and allows 10 clips per video; the Studio plan, at 79 EUR, holds 6,000 credits, 100 hours, and allows 20.
Five steps, the same studio as everyone else.
A published transformation from the community: the original on top, the shorts it became below, with the transcript of each short.
3 of 6 shorts shown · The Override That Saved Humanity · watch it in the community, with the transcript of each short
Demo footage: Tears of Steel © Blender Foundation, licensed CC BY 3.0. Cut, cropped and captioned by shortshort.
No. A shortshort clip is the source cropped to 9:16, with word-timed captions over it and nothing else from the AI: the picture and the sound are the recording's own. A clip is not always one continuous passage, though: to make one point, the AI can join up to 8 non-overlapping excerpts of the same recording, and the studio lists each one with its start and end. The desk can move those bounds, or remove an excerpt, before rendering.
shortshort times every word in a second transcription pass on the selected passage. If the audio does not let that pass line up, the timing from the first pass is kept and the words are flagged as estimated, which tells the desk where to listen again. Captions are in the language spoken and are not translated. There is no audio cleaning, so a noisy recording stays noisy in the clip, and a newsroom with its own verified transcript can import it as an SRT timed to the source.
No. shortshort is driven by hand in the browser: one file uploaded at a time, one MP4 downloaded per clip, with no API or MCP server to call and no analytics; scheduled publishing covers YouTube, with TikTok in early access. Every plan can export a JSON edit plan per clip, with the cut points, the crop keyframes and every word's timing, for a newsroom that renders on its own servers; the Studio plan adds that download for every saved clip.
Count minutes of footage, not clips: one shortshort credit is one minute of source whatever the number of clips. A desk that clips a 40-minute press conference on each of 22 working days needs 880 credits a month, well inside Creator's 1,600; two such conferences a day come to 1,760, 160 above, and Creator at 29 EUR plus a 400-credit top-up at 9 EUR covers it, while the Studio plan's 6,000 credits at 79 EUR a month leaves room for interviews and live captures on top. A weekly 60-minute programme plus a few interviews stays well inside Creator's 1,600 credits. Top-ups on the Creator and Studio plans: 9 EUR buys 400 credits, 24 EUR buys 1,200, 69 EUR buys 4,000, VAT on top.
shortshort analyses the recording from a temporary copy that is deleted once the analysis ends, and the source and the drafts stay private until you export. An exported clip joins the public community by default, with the original video and the transcript of the clip on its page; the Starter, Creator and Studio plans have an account setting that keeps every clip out of it, and any account can take a published clip down, which removes its page from the gallery immediately.
shortshort generates no faces, no voices and no footage: it cuts your recording, reframes it to 9:16 or 4:5 and lays captions over it, so the picture and the sound in a clip are the recording's own. The AI does write text: the proposed title of each clip and, when you schedule one, its title, description and tags, all of which you can edit before it goes out. Each platform sets its own disclosure rules. YouTube's help page asks creators to disclose realistic content that is AI-generated or altered, such as a real person made to appear to say something they did not say, and lists caption creation and an AI-written title among the uses that need no disclosure. This is not legal advice: check the rules of each platform where the desk publishes. YouTube Help: Disclosing use of GenAI content, read .
30 credits cover one video of up to 30 minutes, no card, so start with a shorter recording. Move to Creator when the volume asks for it.