Free AI Tool Turns Text into Videos in Minutes
You can now make custom videos just by describing them—no skills needed.
Someone tried the same "recipe" that produced wild motion-graphics videos with Opus 5.5, but used Qwen3.8 Flash instead — running on an RTX 4080 with 16GB VRAM plus 64GB DDR5 5600, with Strata and Opencode at about 95 tps decode and 3000 tps prefill. The goal: a short video for a nephew who loves trains.
The recipe tells the model to use three.js, Playwright and ffmpeg, then describes the style, dimensions, subject, audio and animation in simple words. The first prompt asked for "a 15-second, 1920×1080, 60fps video in Three.js that showcases the famous japan shinkansen train, like a showreel," rendered frame by frame to MP4, with a 1s hook and a loop.
The first render took about 40 minutes and needed roughly 3 followup prompts to fix rough edges and broken text, for about 2 hours total — the first render was dark, had no buildings, and too much text. According to the article, this threejs + playwright + ffmpeg stack "works pretty well for fixes," and there's no duration limit — videos can be as long as you want. Is it comparable to Opus? "no lol," but "pretty good," and the poster can't wait to try it with Qwen 4.
- A free AI model created a 15-second animated train video from a simple text description.
- The process took about 2 hours on a home computer, with a few follow-up prompts to fix issues.
- This shows AI can make video production cheaper and faster, but quality isn't yet professional-grade.
Why It Matters
Small businesses and creators could save thousands on video production by using free AI tools.