Turn Any Long Video Into 10 Viral Clips.
AI finds the moments people are most likely to watch, share and remember — then turns them into ready-to-post short videos.
Upload your own footage, or paste a direct link to a video you have rights to.
From a two-hour recording to a week of posts
- Step 1
Add your video
Upload a file or paste a direct link. You confirm you have the rights before anything runs.
- Step 2
AI reads the transcript
Speech is transcribed with timestamps, split into sentences, and grouped by topic to find self-contained moments.
- Step 3
The best moments are ranked
Each candidate is scored on hook, curiosity, emotion, story and more. Overlaps are removed.
- Step 4
Clips are cut and captioned
FFmpeg trims, reframes to vertical, burns in captions in your language, and exports web-ready MP4.
Moments are scored, not guessed
Cut points always land on sentence boundaries. Candidates that start mid-thought, end before the payoff, hide long silences or overlap a better clip are pushed down or dropped. The weights below are the defaults, and you can change them.
Default weight of each factor in the viral score
- Hook strength15%
- Story completeness15%
- Curiosity12%
- Emotional intensity10%
- Surprise10%
- Information density10%
- Standalone context10%
- Retention potential10%
- Shareability8%
Captions timed to the word
Word-level timing is used when the transcription provider returns it; otherwise captions fall back to sentence timing. Key words can be picked out in colour.
Post in every language you can reach
The transcript is translated line by line and the timestamps are carried over untouched, so translated captions stay in sync. New languages are one entry in a config file.
English
This changed everything
Hindi
इसने सब कुछ बदल दिया
Hinglish
Isne sab kuch badal diya
Spanish
Esto lo cambió todo
French
Cela a tout changé
German
Das hat alles verändert
Portuguese
Isso mudou tudo
Arabic
هذا غيّر كل شيء
Japanese
これがすべてを変えた
Korean
이것이 모든 것을 바꿨다
Vertical, without the manual crop
Output is 1080×1920 H.264 with AAC audio, optimised for streaming. Reframing runs behind a planner interface: today it centres the crop, and face or speaker tracking can plug in without touching the renderer. Prefer the full frame? Turn reframing off and the video sits over a soft blur.
Questions
Can I paste a YouTube link?
No. ClipForge doesn't download from streaming platforms, since that usually breaks their terms and your rights as a creator. Upload the file you own, or paste a direct link to a video file you have permission to use.
Do I need API keys?
Not to try it. Without keys the app runs in a clearly labelled mock mode: transcripts and scores are placeholders, but uploading, queueing, FFmpeg rendering, downloads and editing are all real. Add keys to switch to real transcription and analysis.
How does it choose which moments to clip?
It segments the transcript into sentences, groups them by topic, builds candidate clips that start and end on sentence boundaries, scores each one on a set of viral factors, then picks the highest-scoring clips that don't overlap.
How long are the clips?
Usually 20 to 60 seconds. You can trim any clip in the editor, from 3 seconds up to 3 minutes.
Where are my videos stored?
On local disk by default, or in any S3-compatible bucket (AWS S3, Cloudflare R2, MinIO). Temporary working files are deleted after each job.
Which languages are supported?
English, Hindi, Hinglish, Spanish, French, German, Portuguese, Arabic, Japanese and Korean, plus Italian and Russian. Adding more takes one line of config.