AI-agnostic selection
Gemini and ChatGPT in the background; Claude, Mistral, DeepSeek, Grok or an AI IDE by copy-paste. Same Master Prompt, same JSON.
Clips
Video → Shorts
One long video → many viral vertical clips with animated word captions.
Paste a YouTube link or drop a file. The app transcribes it on your PC with word timestamps, your AI picks the strongest moments and writes hooks, captions and hashtags per platform, and FFmpeg renders vertical clips with word-by-word captions and a hook title — then you publish or schedule them to YouTube Shorts, TikTok, Facebook and Instagram Reels from the same screen.
- InputYouTube link or video file (mp4, mov, mkv, webm)
- Output9:16 MP4 clips, 1080×1920, captions burned in
- Time1-hour podcast → analyzed in minutes on a GPU
- AIYour Gemini / ChatGPT (hidden browser) or any AI by copy-paste
- Publish toYouTube, TikTok, Facebook, Instagram, Buffer
Who it is for
- Podcasters and interviewers who publish one long episode a week and need daily clips.
- Streamers, lecturers, coaches, preachers — any talking-head content.
- Editors and agencies who clip for several clients and want the metadata written too.
- Creators who tried Opus Clip / Submagic and hit the clip cap, the watermark or the price.
What you get
- 8–15 clips per videoEach 15 s to 14 min depending on the ranges you pick — mix several ranges in one run; "Add more clips" appends a new batch without touching the existing ones.
- Word-by-word captionsThe spoken word is highlighted as it is said. Font, size, colours, position and highlight are per project. Arabic and other RTL languages are laid out right-to-left automatically.
- Hook title + cold openA title pill at the top written by the AI, and optionally a 1–10 s cold-open teaser cut from the strongest line before the clip starts.
- Smart 9:16 framingFace-tracked crop that follows the speaker, or a blurred-canvas layout that keeps the whole 16:9 frame when there are several faces or on-screen text.
- Clean audioFiller words ("um", "uh", "like") and long pauses cut with a threshold you control; profanity bleeped; optional TikTok Community Guidelines moderation.
- Metadata per platformYouTube title, description, tags and category; TikTok caption and hashtags; Facebook / Instagram caption, hashtags and an optional first comment — prefilled in the publish modal.
How it works, step by step
From link to scheduled clips. Every step runs in the background so you can open another project meanwhile.
- 1
New project
Open Video → Shorts and press New project. Paste a YouTube URL (pick the quality; the download runs in the background) or upload a file from your PC. A project card appears at once with live progress.
- 2
Local analysis
faster-whisper (large-v3-turbo) transcribes the audio with word timestamps and voice-activity detection; the result becomes an ID-tagged transcript the AI can reference precisely. Face detection runs when you render. Nothing is uploaded.
- GPU: minutes for an hour of audio. CPU: works, slower.
- Language detected automatically; 99 languages.
- The transcript is stored in the project folder — reopen any time.
- 3
Master Prompt options
Inside the project you set what the AI should find:
- Target durations — presets from < 30 s to 10–14 min; tick several to get a mix; or a custom range in minutes.
- Number of clips — auto or forced.
- Genre — podcast, news, politics, tutorial, vlog, motivational, comedy… tunes the selection criteria.
- Smart Trim — the AI may return boring sub-ranges to cut inside a clip.
- Auto Hook — a hook title per clip and an optional cold-open teaser.
- Auto Censor and TikTok Community Guidelines — ranges to bleep.
- Filler words and pauses removal with a pause threshold.
- 4
Get clips
Three ways: Get clips with Gemini or with ChatGPT runs the prompt in a hidden browser tab of your connected account — several projects at once, extra ones queue, a human check pauses the job and shows you the window. Or copy the Master Prompt into any AI and paste the JSON back. The app validates every clip: IDs resolve to timestamps, durations are recomputed with cuts and teaser, out-of-range clips are flagged.
- 5
Review
Each clip card shows the range, the hook title, a viral score with sub-grades, the AI's reasoning, cleanup stats (fillers, pauses, profanity, cuts, teaser length) and the transcript excerpt. Edit the hook, toggle cleanup per clip, adjust the pause threshold, lock the crop mode.
- 6
Render
Render one clip, queue several, or render all. Face-tracked 9:16 or blurred canvas, word captions with highlight, hook pill, cold open, cuts and bleeps — NVENC when a GPU is available. Renders run in the background across projects; the card shows progress.
- 7
Cloud copy & publish
Finished clips upload automatically to your cloud space (24 h) so publishing works from any PC. The clip row gives Download (with progress), Copy link (24-hour public URL) and Publish: tick accounts, edit the prefilled per-platform texts, publish now or pick a time. Scheduled posts fire from the cloud even when the PC is off.
Features in depth
Everything a paid clipper does, plus what it does not.
Multi-range runs
Ask for < 30 s clips and 3–5 min clips in the same run; the AI returns both and the validator checks each against its own bucket.
Cold-open teaser
The strongest sentence plays first with the hook title, then the clip restarts from its natural beginning — the format that stops the scroll.
Face tracking
YOLOv8 finds the speaker; the crop follows smoothly. Multiple faces or text on screen switch to the blurred 16:9 canvas so nothing is cut off.
RTL captions
Arabic, Hebrew, Persian and Urdu captions are shaped and aligned right-to-left, word highlight included.
Brand-safe audio
Profanity bleeps in several languages and an optional TikTok Trust & Safety pass that flags any guideline violation, not just swearing.
Scheduling from the cloud
Pick a date and time per post; the cloud scheduler publishes while your PC sleeps; status badges update live on the clip.
Works across PCs
Rendered clips live 24 h in your cloud space: publish from a laptop what you rendered on the desktop.
Background everything
Downloads, analyses, AI runs, renders and uploads continue when you leave the page or open another project.
Every setting you control
Nothing is hidden behind a paywall or an "advanced" plan — these are the knobs in the app.
| Setting | Options | Notes |
|---|---|---|
| Target durations | < 30 s · 30–59 s · 60–89 s · 1.5–3 min · 3–5 min · 5–10 min · 10–14 min · custom minutes | Tick several for a mix. TikTok API caps third-party posts at 10 min. |
| Number of clips | Auto · forced count | Auto lets the AI decide from the content. |
| Genre | Auto detect · News · Politics · Podcast · Tutorial · Vlog · Motivational · Comedy · … | Changes the selection criteria in the prompt. |
| Smart Trim | On / Off | AI returns sub-ranges to cut inside a clip. |
| Auto Hook / Cold open | On / Off · teaser 1–10 s | Hook title pill + optional cold open. |
| Auto Censor · TikTok Guidelines | On / Off | Bleep ranges from the AI; guideline pass is stricter than profanity. |
| Filler words · Pauses | On / Off · pause threshold 0.5–5 s | Per project defaults, per clip overrides. |
| Captions | Font · size · colours · highlight · position · on/off | Windows fonts and downloaded Google fonts. |
| Framing | Auto · face crop · blurred canvas | Lockable per clip. |
| Encoder | Auto (NVENC → CPU) | Shown in the app header. |
| Output folder | Any folder | Settings → Output location; per-tool subfolder video-to-shorts/. |
What you need
- A Windows 10 / 11 PC; NVIDIA GPU optional but recommended
- A YouTube link or a video file
- A Gemini or ChatGPT account connected in Integrations — or any AI for copy-paste
- Linked accounts to publish (YouTube, TikTok, Facebook / Instagram) or a Buffer key
Tips & best practices
- Talking-head content works best: podcasts, interviews, sermons, lectures, monologues.
- Mix a short bucket (< 30 s) with a mid bucket (60–89 s) in one run — the AI picks different moments for each.
- Turn on the cold open for storytelling episodes; leave it off for tutorials where the first sentence is already the hook.
- Use "Add more clips" instead of re-running: it appends a fresh batch and never touches rendered clips.
- Let clips upload before closing the app — the publish modal needs the cloud copy.
Honest limits
What the tool does not do (yet), so you are not surprised.
- Windows only; no macOS / Linux build yet.
- Music videos and content without speech give the AI little to work with — the selection is speech-driven.
- TikTok third-party posting is capped at 10 minutes by TikTok; longer clips can still go to YouTube and Facebook.
- Cloud copies expire after 24 hours; render again (or re-upload) to publish later.
Frequently asked questions
How many clips do I get from one video?
Typically 8–15 per hour of speech, depending on the durations you tick and the content. You can force a count or run "Add more clips" for another batch.
Does the AI see my video?
No. It receives the ID-tagged transcript and your settings as text; the app does the cutting locally from the timestamps.
Can I edit a clip before rendering?
Yes: hook title, cleanup toggles, pause threshold, crop mode. Start / end come from the AI's IDs; ask for another batch or a fix if a range is wrong.
Which caption styles are there?
Word-by-word with highlight of the spoken word; you choose font (Windows or Google fonts), size, colours, highlight colour and position per project.
Does it publish to Instagram?
Yes — Instagram professional accounts linked to your Facebook pages get the clip as a Reel with caption, hashtags and optional first comment.
Is there a watermark or a monthly limit?
No watermark, no limit. The only cost is your time and electricity.
Clip your first video
Free download for Windows. A one-hour podcast becomes a week of Shorts in one sitting.
Free · No watermark · No clip limits · Runs on your PC