Manul

Direct your video. Manul edits it.

Drop in your footage, say what you want, and point at the frame when something's off.
Manul makes the cut on your own computer and shows you a before / after of every change.
You never touch a timeline.

Get Manul · Roadmap · Contribute

https://github.com/user-attachments/assets/084276fb-6129-47e2-9cba-4724c1432990

_{90 seconds with Manul, starring the grumpiest cat on Earth. Pallas's cat footage: BBC, Frozen Planet II.}

The Manul editor: the film, notes on the scrubber, the agent's steps, a note pinned to a box on the frame, and a before/after proposal

Why it's different

Video editors make you operate them. AI video tools take your footage away and hand back whatever they made. Manul works the way a director works with a good editor in the room.

You point. It edits.

Draw a box around a number plate that needs blurring. Drag across the sentence that drags on. Click the one word in a title that's wrong. Your note sits on that exact spot and moment, and that's the spot Manul changes.

It proposes. You decide.

Nothing is overwritten. Every edit arrives as a new version you can scrub before and after, then keep or send back with a reply. Every version you ever kept is one click away.

Your pixels stay real.

Manul cuts your footage. It never swaps your shots for generated ones: generated b-roll is coming, but only where you ask, labelled as generated. Titles, lower thirds and captions are real typography, drawn frame-perfect, and you can still drag the words around after Manul has made them.

Your footage stays home.

No upload queue, no cloud project, no waiting for a render farm. The editing and even the transcription run on your computer. The AI you bring your own key for only ever sees text and still frames.

What a session feels like

The start screen: drop a video and say what you want, start from an idea, or open a recent film

You: cut the ums, and the bit where I lose my train of thought

Manul: Cut 9 fillers and the restart at 0:42. 2:31 → 2:12. before / after Keep · Send back

You: (draws a box over a parked car, 0:18–0:24) blur this number plate

Manul: Blurred that box from 0:18 to 0:24. The rest of the shot is untouched.

You: put my name on screen when I introduce myself, and add quiet music under the whole thing

Manul: Lower third "Sara Malik · Product designer" at 0:04. Music looped to 2:12, dipping under every sentence.

You: now a vertical one for Shorts, captions burned in

_{An example of the kind of conversation Manul is built for. Every reply comes with a before / after you can scrub.}

What's in the box

	Word-accurate cuts Every spoken word is transcribed with its timing, so "cut the ums" or "drop the second take" lands on the exact word
	Titles it designs Title cards, lower thirds, callouts and animated text, timed to your film and editable by hand
	A mix you hear live Music loops to the length of your film and ducks under speech. Move a slider and you hear it at once, before anything renders
	One film, every shape 16:9 for YouTube, 9:16 for Shorts, Reels and TikTok, 1:1 for feeds, with captions burned in or as a file
	Its own browser It looks up references and finds footage in a browser of its own, never yours, unless you switch that on
	A memory for taste Say "captions are always yellow" once and every project after that knows it
	Tabs and any model Several films open at once, each with its own conversation. Claude, Gemini or GPT, picked per project

You stay in charge

Settings, Browser: Manul's own browser by default, your Chrome only if you choose it

Which browser it uses
_{Its own, by default. Yours only if you say so.}

Settings, Tools: built-in tools, speech recognition found on this computer, heavier tools downloaded only when needed

What it installs
_{Nothing heavy until you ask, and it uses what you already have.}

Get Manul

Manul is in early days.

brew install hormuz-labs/tap/manul     # macOS

Or hand it to your coding agent (Claude Code, Cursor, Codex…) and let it do the install:

Install Manul from https://github.com/hormuz-labs/manul. On macOS run `brew install hormuz-labs/tap/manul`. On Debian or Ubuntu, add the apt repository at https://apt.manul.si (the three commands are on that page) and run `sudo apt install manul`; on other Linux, download the AppImage for this machine's architecture from https://github.com/hormuz-labs/manul/releases/latest. Then open Manul.

On Debian or Ubuntu, add the signed apt repository once and apt keeps Manul up to date:

curl -fsSL https://apt.manul.si/manul.gpg | sudo gpg --dearmor -o /usr/share/keyrings/manul.gpg
echo "deb [signed-by=/usr/share/keyrings/manul.gpg] https://apt.manul.si stable main" | sudo tee /etc/apt/sources.list.d/manul.list
sudo apt update && sudo apt install manul

Other Linux: grab the AppImage from the latest release. More at manul.si.

It's free and open source. Bring a Claude, Gemini or OpenAI key (it's kept in your system's keychain), or wait for the single Manul key, coming soon.

Want to hack on it? Run it from source with CONTRIBUTING.md.

Where it's going

Now	Directing by talking and pointing, titles, music, every export shape
Next	Hand the cut to Premiere, Resolve or Final Cut · publish straight to YouTube
Then	One Manul key for every model · heavy tools in the cloud · your apps connected
Later	Shoot on your phone straight into a project · clients leave notes on the frame · teams

See the full roadmap →

_{Named after the manul, the Pallas cat: small, patient, and unbothered by anything.}
_{GPL-3.0 · Third-party notices · Contribute · Made by Hormuz Labs}