How I Make AI Shorts for About a Dime in Voice Costs

This video was made by AI. The script, the voice, every animation. Here's the whole stack I use for my Shorts, what each piece costs, and how you can build the same thing.

What happened

This one is about my own pipeline, so the facts come from how I actually build these, plus the vendors' public pricing.

Step one: find the story. Claude scans Hacker News and other sources for what's trending, then opens the original source and checks every claim before anything goes in a script. If a fact can't be confirmed, it gets cut.

Step two: write the script. Claude drafts several hooks, scores them, and writes the script in my style. Every Short opens with the hook on screen in the first frame.

Step three: the voice. ElevenLabs reads the script in my cloned voice and returns a timestamp for every single word. Those timestamps drive everything that follows.

Step four: the animation. Remotion, which calls itself a way to "make videos programmatically" with React, animates each scene on the exact word I say. Captions, images and sound effects all land on the word timings from step three.

What it costs. ElevenLabs lists its Creator plan at $22 a month for 121,000 credits, with a first-month discount, and text to speech costs one credit per character on its standard models. A Short script runs about 600 characters, which works out to roughly ten or eleven cents of voice. Rendering happens on my laptop, so it costs nothing extra, and Remotion is free for individuals and teams of up to three people.

I want to be straight about the full picture. The dime covers the voice on the basic stack. When I add AI-generated images and a short generated video clip for the hook, as I do on most Shorts now, that adds a little over a dollar per video in image and video credits. Still cheap, just more than a dime.

Why it matters for builders

Short-form video used to need a camera, an editor and a few hours per clip. This stack makes one in minutes once it's set up, and every part of it is code you control.

Because the edit is a React project, you can version it, template it and change it with an AI coding agent. Want a new caption style across every future Short? That's one component change. Want a branded intro for a client? That's a new composition you reuse forever.

The word-level timestamps are the piece most people miss. Without them, you're dragging clips around a timeline by hand. With them, animation lines up with speech automatically.

Try it in 15 minutes

1. Install Node.js, then create a project: npx create-video@latest. Pick a starter template and run the preview.
2. Make an ElevenLabs account. The free plan includes 10,000 credits a month, enough to experiment.
3. Write a 20-second script with the hook in the first line.
4. Generate the voice through the ElevenLabs API using the text-to-speech endpoint that returns timestamps, and save both the audio file and the timing data.
5. Load the audio into your Remotion composition and show each word as a caption at its start time.
6. Render it locally and watch it on your phone.

Then ask Claude Code to add one scene change on a keyword. That's the loop you'll repeat to build out your whole style.

Here's how you get paid

My opinion: businesses already pay agencies every month for short-form video, and this stack makes one in minutes.

That's a content agency where the only employee is you. Here's how I'd set it up:

• Build one template per client. Their colors, fonts, logo and caption style, all in a Remotion composition you reuse.
• Clone a voice only with permission. Use the client's own voice with their written consent, or a licensed stock voice.
• Sell a monthly package. A set number of Shorts per week, with research, scripts, voice and edit included.
• Label AI use. Turn on the platform's AI disclosure every time. It protects you and your client.

Your margin comes from the template. The first video for a client takes a while. The fiftieth takes minutes.