Auteur: auto edit reels with our AI editing engine

makerzz · AI editing engine

Our AI editing engine — the makerzz editor. It reads your take, plans the cut, builds visuals on every beat, captions it and checks every scene before you see it.

Edit a video with Auteur What's new

makerzz.space/studio/projects/your-take

Auteur · done

6/6 scenes ready

Ready to review

01
Hook: Claude can run your channelready
02
What it handles, ticked offready
03
The repo, on its real pageready
04
11 skills, one channelready
05
Typing the one commandready
06
It writes the whole scriptready

Ready

A real reel in the Glitch style, replayed scene by scene. The frames are from the finished video.

How it edits

Not one prompt. A team of agents.

Each part of an edit has its own agent with one job. They hand work to each other the way an editing team does, and every scene is checked before the final render.

1/7

Planner

Reads your transcript and splits it into scenes, about one per spoken sentence. For each scene it decides what should be on screen, which real material it needs, and where your face shows.

Illustration

Scene plan

01Hook: Claude can run your channel0:00 · one spoken sentencesplit-hook
02What it handles, ticked off0:03 · one spoken sentencesteps
03The repo, on its real page0:06 · one spoken sentencegithub-card
0411 skills, one channel0:09 · one spoken sentencecounter + grid
05Typing the one command0:13 · one spoken sentenceterminal
06It writes the whole script0:16 · one spoken sentencesteps

2/7

Scene agents

One agent per scene, all working at the same time. Each one writes its scene as code against the house style, using your screenshots, logos and the sources it gathered — not stock clips.

Real frames

Six agents, six scenes, same minute

agent 1
agent 2
agent 3
agent 4
agent 5
agent 6

Frames from the finished Glitch reel above.

3/7

Style engine

The part we built ourselves. Every house style is a set of measured rules — layout, timing, type, motion curves, sound — so a scene written today moves exactly like the reference it was studied from.

House style: Glitch

Rules the scene agents write against

Avatar card: bottom 45% of the frame
Avatar rhythm: ~4 s on, ~6 s of graphics
Captions: 1–2 words, fixed low
Entry / exit: rise in, whip out, hard cut
Something lands: on every key spoken word
Nothing still: for more than 0.6 s

4/7

Checker

Renders each scene on its own and checks it before the full render: does it run, is the frame full, is the type readable, are the real assets used, does the screen match the source.

Per scene, before anything is rendered

Every scene renders without an error
The frame is filled, not a small island in the middle
Type is big enough to read on a phone
Real screenshots and logos are used where they exist
What's on screen matches the source, mismatches are flagged

Fails twice? The scene agent rewrites it, then falls back to a simpler, proven layout so the video still ships.

5/7

Renderer

Renders the final frames in parallel segments, then mixes your voice, one music bed and a sound effect on each visual event, mastered to a loudness level that sits well on every platform.

A finished reel in the Graphics First house style — Graphics First

6/7

Captioner

Writes the on-screen captions from your exact words, then a caption for every platform you post to: an Instagram caption, a YouTube title, a LinkedIn post. Add a keyword and it writes the “Comment …” line too.

Illustration

One video, a caption per platform

InstagramComment SKILLS and I'll send you the setup 👇 One command now writes my whole YouTube script…
YouTubeTitle: 11 Claude skills that run a YouTube channel #shorts
LinkedInI handed my YouTube workflow to 11 Claude skills. Here's what it does well, and the one thing it can't.

7/7

Publisher

Once you approve, it posts or schedules to the accounts you pick and switches on the comment-to-DM flow with the post. You review every video before anything goes out.

Illustration

After you approve

InstagramToday, 6:00 pmScheduled · DM flow on
YouTubeTomorrow, 12:00 pmScheduled
TikTokTomorrow, 9:30 amScheduled

Someone comments SKILLS → they get the setup in their DMs.

Style engine

Built from measured motion

We took reference reels apart frame by frame and reduced every move to a curve. The engine plays those curves back, so scenes enter, travel and leave the way a top editor's do — not with a generic ease.

Rise

2 s

190·e^(−t/0.5) + 70·e^(−t/0.08) px

How a scene enters: it comes up from 260 px below, fast at first, then eases in for a long soft landing.

Whip

233 ms

−2205·τ² px, last 7 frames, then a hard cut

How a scene leaves: it accelerates up and out over the final 7 frames (about 120 px) and the next scene cuts in.

Settle

2.5 s

two-pole exponential, τ = 0.5 s

How the camera moves on a spoken word: it starts gently, travels, and settles with a half-second tail instead of stopping dead.

Measured frame by frame on a reference reel with phase correlation and tracking, then rebuilt in code and compared side by side (SSIM on the stage area). Every house style is built the same way: real measurements first, then the engine.

Multilingual

Speaks your language

Record in the language you talk to your audience in. Auteur transcribes it as spoken, and the captions and on-screen text come back in that same language — Hinglish stays Hinglish.

EnglishThree tools that replaced my editor
HinglishTeen tools jinhone mera editor replace kar diya
हिन्दीतीन टूल जिन्होंने मेरा एडिटर बदल दिया
EspañolTres herramientas que reemplazaron a mi editor
PortuguêsTrês ferramentas que substituíram meu editor
DeutschDrei Tools, die meinen Cutter ersetzt haben
FrançaisTrois outils qui ont remplacé mon monteur
Bahasa IndonesiaTiga alat yang menggantikan editor saya

One caption line in eight of the languages you can record in. Examples, written for this page.

Long-form

Raw footage in, finished video out

Upload the whole session. Get the clean cut.

Drop in one or several raw files — up to an hour. Auteur finds the retakes, false starts, long silences and filler, keeps the best take of each line, and hands back a clean 16:9 cut.

Every cut is listed with the words and the reason
Chapters and captions for the long video
3–5 shorts pulled from it, reframed to 9:16
Add a graphics pass in a house style if you want one

Illustration

Your raw footage, decided line by line

Kept
Retakes
Silences
Filler
Off-topic

“A valve the size of a coin… no, again.”

“On the way home, a valve the size of a coin shook open.”

Hover any segment in your project to read what was said and why it was cut.

House styles

One engine, 26 looks

Each house style is a full set of rules for the engine. Pick one per video, or let it edit the same take in a few and choose.

All styles

What it runs on

Auteur runs on leading third-party AI models plus our own style engine, checks and render pipeline. The models do the reading and writing; the house styles, the scene plan, the checks, the motion and the rendering are ours. We did not train our own foundation model. You review every video before it is posted.

Curious how we will measure it? Read the benchmark method.

Auteur

Give it your next raw take

Start free with your first edited video. No card needed.

Edit my first video