Free · Open source · Self-hosted
Type a topic.
Get a Short.
TubeMatic makes vertical explainer videos on your own machine. Give it a topic and your AI coding agent researches it, writes the script and animates every scene. ElevenLabs voices it, and the result is ready for YouTube Shorts, Reels and TikTok. You approve the topic, the script and the finished video, or turn on Auto Pilot and let it run by itself.
How does a packet find its way across the internet?
TubeMatic makesWhat happens to your topic
waits for your approval
- TopicAgentYou approve
- ScriptAgentYou approve
- VoiceElevenLabs
- ScenesAgent
- RenderRemotion
- ReviewYouYou approve
- PostYou
What it does
One Short on request, or a steady supply
Ask for one video, or let TubeMatic keep finding topics and line them up for your approval.
Where topics come from
Ask for one video
Type a title and a topic, pick a voice and a look, and it goes straight to scripting.
Run a series
Seed a subject like “Python internals” and TubeMatic proposes episode after episode for you to approve.
Follow what’s trending
It scans GitHub Trending, Hacker News, Lobsters and Dev.to for topics suited to a short explainer.
Start from a long video
Point it at YouTube videos and it pulls out the ideas they only mention in passing. Each idea can become a Short of its own.
What you control
Write in a style you like
Give it YouTube videos whose writing you like. It scripts new videos in their style, without reusing their content.
Post now or later
Publish a finished Short to YouTube, Instagram, TikTok and Facebook from one dialog, right away or at a date and time you pick.
Run it hands-off
Turn on Auto Pilot for the projects you choose. It approves topics, scripts and videos by itself and posts to the accounts you pick, with the spacing you set.
Track views and spend
See views, likes, comments and shares for every post alongside what the video cost, broken down by provider and model.
The look
Three looks, or leave it on Auto
Choose a style when you approve the topic. Every style gets a voiceover and captions that light up word by word. These frames come from Shorts made with TubeMatic.

Motion Graphics
A dark canvas with generated imagery and animated diagrams.

Stick Explainer
Black line art on white. Its figures play roles, like “the user” or “an attacker”, which suits a walkthrough or a what-if.

Illustrated
Generated characters stay recognizable from scene to scene, which suits a story that follows one person.
Auto
No fixed style. TubeMatic picks a dark or a light canvas for the whole video, then gives each beat what its subject needs: a generated image, a diagram or a figure.
How it works
From a topic to a posted Short
The agent does the work, and you decide at three points: whether a topic is worth making, whether the script is worth paying to voice, and whether the finished video is ready to post.
- Agent
Find a topic
Topics come from a series you seed, from what’s trending, from YouTube videos you point it at, or straight from you. Researched topics are scored from 1 to 10 for how well they suit a short explainer, and TubeMatic turns away anything under 5.
- You
Approve the topic
Approve it, reject it or ask for changes. This is also where you pick the voice, the look and the script profile.
- Agent
Write the script
The agent writes it beat by beat: the words to speak, and a brief for what’s on screen while they’re spoken.
- You
Approve the script
Read the script before it goes to ElevenLabs, then approve it or ask for changes.
- ElevenLabs
Voice it
ElevenLabs reads the script in the voice you chose, and every word gets a timestamp so the captions land on the beat.
- Agent
Build the scenes
The agent generates the imagery with OpenAI and writes a React component for every scene. The whole project is typechecked before a single frame renders.
- Remotion
Render
Remotion and Chromium render a vertical 1080×1920 mp4 at 30 fps, usually 90 to 180 seconds long, with the captions burned in.
- Gemini
AI review, if you turn it on
Gemini watches the finished video, not the script, looking for defects, and anything it finds goes back to the agent to fix before the video reaches you. It’s off by default, still experimental, and needs a Gemini key.
- You
Approve the video
Watch it in the dashboard, then approve it or send it back with notes.
- You or Auto Pilot
Post it
Download the mp4, or publish it yourself right away or on a schedule. With Auto Pilot on, TubeMatic posts each finished video for you, within the daily limit and the gap between posts you set for each account. Publishing is off until you turn it on.
Want it hands-off? Turn on Auto Pilot for the projects you pick. TubeMatic then approves its own topics, scripts and videos, and posts them on the spacing you set. It’s off until you turn it on.
The dashboard
Every decision in one place
You run TubeMatic from a dashboard in your browser, served from your own machine. The Pipeline page puts every pending decision at the top, where you can read a script, watch a video and approve it or send it back.

- Overview
- Counts of what’s waiting on you, what’s in progress, what’s published and what failed.
- Quick actions
- Every topic, script and video that needs a decision, with Watch, Approve and Revise.
- Kanban
- Each video moving through the pipeline, one column per status.
- Notifications
- A Discord message tells you when a topic, script or video is waiting for you.
Under the hood
Every frame has source code
The agent writes a React component for each scene of each video, instead of filling in a template, and Remotion renders it. When a review sends a video back, the agent edits that code. You can read the source that produced any frame.
import { AbsoluteFill, Easing, interpolate, useCurrentFrame } from "remotion";
import { Captions, Node, Packet, Wire } from "../components";
// Written by the agent for this one video.
const CLIENT = 0.27, SERVER = 0.65;
const packets = [
{ label: "SYN", from: CLIENT, to: SERVER, at: 150 },
{ label: "SYN-ACK", from: SERVER, to: CLIENT, at: 200 },
{ label: "ACK", from: CLIENT, to: SERVER, at: 250 },
];
export const Handshake: React.FC = () => {
const frame = useCurrentFrame(); // 320
const open = frame >= 300;
return (
<AbsoluteFill className="bg-ink">
<Node label="Client" y={CLIENT} appearAt={10} />
<Node label="Server" y={SERVER} appearAt={40} />
<Wire from={CLIENT} to={SERVER} drawFrom={84} solid={open} />
{packets.map((p) => {
const t = interpolate(frame, [p.at, p.at + 45], [0, 1], {
easing: Easing.inOut(Easing.cubic),
extrapolateLeft: "clamp",
extrapolateRight: "clamp",
});
return <Packet key={p.label} label={p.label} y={p.from + (p.to - p.from) * t} />;
})}
<Captions timeline="timeline.json" />
</AbsoluteFill>
);
};
How TCP says hello
A scene written the way TubeMatic writes them, running on the same frame as its source.
What you need
Your own machine and API keys
TubeMatic is free. It needs a Linux machine with Docker, an AI coding agent you already use and API keys for the providers below. The install wizard asks for each key, checks it with the provider and starts everything.
- Agent
- Claude Code, Codex or OpenCode. Install one or several, and share the work between them.
- Runs on
- Your own Linux machine, in Docker. One install wizard sets it up. It needs at least 2 CPU cores, 4 GB of memory and about 15 GB of disk.
- Voice
- ElevenLabs, with your key
- Images
- OpenAI, with your key and billing enabled. It draws every scene image and the Illustrated characters.
- Optional keys
- Gemini for topic mining and AI review. youtube-transcript.io for script profiles. postforme.dev for publishing.
- Render
- Remotion + Chromium
- Output
- 1080×1920 mp4 · 30 fps · 90–180 s · captions burned in
- Posts to
- YouTube · Instagram · TikTok · Facebook, through postforme.dev. By hand, on a schedule or with Auto Pilot. Optional.
- Price
- Free. You pay the providers you use, and the dashboard shows what each video cost.
- License
- Apache 2.0. Remotion, which renders the videos, has its own license: free for individuals, non-profits and companies with up to three employees, and paid for larger companies.
Made with TubeMatic
See more of what TubeMatic makes
Two Instagram channels run by TubeMatic’s developer, posting Shorts made with TubeMatic.
Free and open source.
Now on GitHub.
TubeMatic is released under the Apache 2.0 license. The quickstart takes you from a Linux machine with Docker to your first video.