How do I turn a long video into YouTube Shorts with AI?

Why this happens

How to do it step by step (no special tool needed)

This works with any editor and a free command-line tool. Plan on one to two hours for the first episode and less once you have a routine.

1. 1. Start from a file you own

   Use your own podcast, webinar, tutorial or stream, ideally the original export rather than a compressed download. YouTube's reused-content rule says it applies even if you have permission from the original creator and is separate from copyright, so clipping other people's videos is a channel-level risk, not just a copyright one.

2. 2. Get a transcript with timestamps

   Reading is faster than scrubbing. Export an SRT or WebVTT transcript from whatever tool you already use and fix names and technical terms, because errors end up in your captions. If you only have the audio, any speech-to-text tool that outputs timestamps will do.

3. 3. Mark the moments worth cutting

   Choose stretches that pass three tests: the thought starts and ends inside the clip, the first sentence is a hook (a question, a claim or a specific), and it makes sense to someone who has not seen the episode. Aim for three to eight per hour of video. The clip-picking checklist goes deeper.

4. 4. Cut on sentence boundaries

   Set the start just before the first word of the hook and the end just after the last word of the closing sentence. Put -ss before -i so ffmpeg seeks quickly, and re-encode so the cut lands where you asked.

   ffmpeg -ss 00:12:03.4 -i episode.mp4 -t 47.6 -c:v libx264 -crf 20 -c:a aac clip01.mp4

5. 5. Reframe to 9:16

   For one speaker, crop a vertical strip from the middle of the frame and scale it to 1080x1920. If the speaker sits off-centre, change the x offset by hand. For two people, stack two crops (one above the other). A fixed crop does not follow a speaker who moves.

   ffmpeg -ss 00:12:03.4 -i episode.mp4 -t 47.6 \
     -vf "crop=ih*9/16:ih:(iw-ih*9/16)/2:0,scale=1080:1920,setsar=1" \
     -c:v libx264 -crf 20 -c:a aac clip01_vertical.mp4

6. 6. Add captions

   Trim your corrected transcript to the clip, shift the timestamps so the clip starts at zero, and burn the captions into the picture so they show on every platform. Keep to two lines and lift them off the bottom edge so app buttons do not cover them.

   ffmpeg -i clip01_vertical.mp4 \
     -vf "subtitles=clip01.srt:force_style='Fontsize=18,Alignment=2,MarginV=300'" \
     -c:a copy clip01_captioned.mp4

7. 7. Review each clip, then upload

   Watch every clip once with sound, as someone who has never seen the source. If you need the sentence before it to follow, start earlier. Then upload from YouTube Studio or the YouTube app. Square or vertical files up to three minutes are categorised as Shorts. A Short over one minute with an active copyright claim is blocked, so keep claimed music out of longer clips.

Our own summary of the workflow, not a benchmark: we have not timed either route.

When to use RemixViral

RemixViral automates steps 2 to 6 for a video you upload yourself (MP4, MOV, MKV or WebM, up to 120 minutes). It transcribes the whole video, ranks the candidate moments best first with a one-line reason and a score for each, cuts them to 9:16 with captions burned in, and shows them in your dashboard for review. You choose how many clips (5, 10, 15 or 20), their length and caption style.

Ask your AI assistant to do it

Once you have uploaded a video and the clip job has finished, you can hand the review to your AI assistant over MCP and keep the decision for yourself.

Assign to RemixViral: use list_clip_jobs to find my latest finished clip job, then get_clip_job to read its ranked clips. Give me a table of the top five with start and end times, score and the reason each was picked. Do not create anything or spend credits. Show me what you did so I can approve which clips to export.

How to connect RemixViral to Claude, ChatGPT, Cursor or Gemini, and the full tool list: AI assistants (MCP).

Frequently asked questions