For

Vyra for YouTube creators: edit real footage by describing the edit

Vyra turns hours of raw footage into a finished long-form video or a batch of Shorts by describing the edit. It reads transcripts and scenes, so name moments.

By Sulan Zhang, co-founder of Vyra · Updated 2026-09-24

Vyra edits your real footage from a plain description. It transcribes and analyzes everything you upload, so you can say "cut the best takes into a 10-minute video and pull three Shorts from it" and get a timeline you can still adjust by hand.

A real week with Vyra

One Saturday shoot, 38 minutes of talking head plus 20 b-roll clips. Goal was a 14-minute video and three Shorts from the same footage.

Cut this into a 14-minute video. Keep the best take of each point, remove pauses over half a second and restarts, keep my section order. Cut to the matching b-roll whenever I name a tool. Title card at each section. Once that's locked, pull three Shorts under 50 seconds each, 9:16, me centered, word-by-word captions, one-line title on top for 2 seconds.

The long cut came back at 14:20 with title cards using the section names from the transcript. Two b-roll picks were wrong, so "at 6:12 use the desk shot, not the street" fixed both. The three Shorts used the cleaned takes, so nothing needed re-cutting. Long video exported at 1440p Sunday, Shorts posted across the week.

What you make

Your three most-used prompts

Rough cut of a long video

Cut this 40-minute recording into a 12-minute YouTube video. Keep the best take of each point, remove pauses over half a second, retakes, and filler words. Keep the section order I filmed in. No captions, no music yet.

Pull Shorts from the long edit

Find the three strongest self-contained moments under 50 seconds each. Make a 9:16 version of each with me centered, word-by-word captions in bold white, and a one-line title at the top for the first 2 seconds.

Chapters and lower thirds

Add a title card at the start of each main section using the section names from the transcript. Add a lower third with my name and handle for the first 5 seconds. Simple slide in from the left.

A typical workflow

Upload the raw recording and any b-roll. Vyra transcribes every word and analyzes every scene.
Ask for a rough cut with a target length and cut rules.
One correction per message. Swap a take, tighten a section, fix a caption.
Ask for the Shorts once the long edit is locked, so they inherit the cleaned takes.
Export at 1440p on AI Starter or 4K on AI Pro, then export the 9:16 Shorts.

What Vyra does that matters for you

It searches your footage by what was said and what's on screen. "Find every time I show the graph" works across an hour of tape.
It removes filler and retakes with rules you set, which is where most of a YouTube edit's time goes.
It reframes 16:9 to 9:16 with masks that follow you, so Shorts come from the same project.
The full timeline is always there for the last 10 percent.

What it does not do

It doesn't generate footage. Everything on the timeline is something you uploaded.
It doesn't upload to YouTube for you.

Example

Example creator (TODO)

FAQ

Can it handle a 2-hour recording? Yes. Upload is analyzed once. After that, prompts about specific moments use the transcript and scene index.

Will it cut jokes or pauses I meant to leave in? Only if you let it. Say "don't touch 4:10 to 4:25" or "allow pauses up to one second."

Do I need an AI subscription? No. The built-in chat works on its own. If you already pay for Claude, ChatGPT, or Cursor, you can connect it through MCP instead.