Open-source AI video editor

You bring the raw footage.
It does the rest.

Drop in a recorded talk, photos and a logo, and say what you want. A team of AI agents researches, cuts, animates and checks every frame, then hands you a finished, on-brand video. It runs sandboxed on your computer, with your own Gemini key.

Install in one line View on GitHub
Free & open source (MIT)Mac · Windows · LinuxAbout $1–4 per video
localhost:4322 · Elyps AI
From this, to this

One message in.
A finished explainer out.

The raw interview footage The edited frame: the speaker beside an animated graphic Raw footageElyps edit
The whole brief“Turn this interview into a short explainer for our channel, Earth, Explained. Add graphics at the key ideas, a clean intro and outro, and soft music. Use our logo and its colours.”
How it works

A lead agent,
and a team of subagents.

It works like a small studio. The lead agent understands the material and plans the edit with you. Then it hands each graphic, and any research, to its own subagent, and they all work at the same time.

1

Understands everything

Syncs the camera and the mic, transcribes every word, pulls your brand's colours from your logo or website, and looks up the speaker.

2

Plans with you

Sends a beat sheet: what's said when, and the graphic for each idea. You approve it or redirect it, in plain words.

3

Builds in parallel

Subagents research the web, capture pages, and write each animation as code, all at the same time.

4

Checks its own work

Every graphic is rendered, looked at and fixed before it goes in. Then the lead renders the final 1080p video.

Lead agentPlanning the edit
What it does

Everything an editor does.
Without the timeline.

Cuts, syncs and captions

Lines up a separate mic with the camera, trims the silences, and times everything to a word-level transcript.

Animated graphics, written as code

Charts that draw themselves, diagrams that build as they're named. Each graphic is its own HTML, SVG and GSAP animation, timed to the words.

Researches the web

Looks up the speaker and the claims in the talk, with sources. It captures pages as screenshots or smooth scrolling recordings, with the key line highlighted.

Illustrations and music

Draws illustrations in your brand's colours and composes a music bed that dips under the voice.

Change anything

Click a graphic on the timeline and say what's different. Or message the agent while it works; it picks the note up at its next step.

Sandboxed, and your key stays yours

The agents run inside a container on your computer, as their own user. They reach Gemini through a local proxy and never see your API key.

Frames from Elyps's edit of an interview with Dr. Lesley Ott, courtesy of NASA's Goddard Space Flight Center. “Earth, Explained” is a made-up channel for the example. NASA does not endorse Elyps.

Install

One line. Then drop in a video.

Elyps runs in Docker Desktop, so everything it needs comes in one package. Give Docker 8 GB of memory in its Settings → Resources.

curl -fsSL https://raw.githubusercontent.com/ayushpatnaikgit/Elyps-AI/main/install.sh | sh
  1. It opens Elyps in your browser.
  2. Add a Gemini key from aistudio.google.com/apikey when it asks.
  3. Drop in your footage and say what you want.
A full edit of a 3–4 minute talk≈ $1–4
Each illustration$0.07
Each music track$0.08

You pay Google directly for Gemini usage; Elyps itself is free. The Usage page shows what each video cost.

Elyps AI

You bring the raw footage. It does the rest.

Install Elypsgithub.com/ayushpatnaikgit/Elyps-AI