The Viral Voice Synth Called CHIMERA Isn't AI. It Simulates a Throat.

CHIMERA, a free browser voice synthesizer showing up in clips captioned as AI, has no neural network. It simulates air moving through vocal folds, a vocal tract, lips and nose, and gathered about 1,900 GitHub stars in six days.

Reporter bylines on BizTechDaily are editorial personas, and articles are drafted with AI tools. The site is run by its publisher, who writes as Hank Reed. How we report

The Viral Voice Synth Called CHIMERA Isn't AI. It Simulates a Throat.

AI-generated illustration

CHIMERA, the voice synthesizer turning up in short videos this week, does not use a neural network, according to its GitHub repository. "No samples, no neural audio," the README says. Every sound comes from a simulated airway of lungs, vocal folds, a branched vocal tract, lips and nose. Some viral posts have called it AI anyway, including a DramaAlert post saying "AI is now able to simulate human vocal cords."

The engine is written in Rust, compiled to WebAssembly and run in the browser, simulating the airway at 96 kHz for an adult human voice, per its architecture notes. Users drag the tongue, lips and jaw in a cutaway head, type English text or phonetic symbols to make it speak, and write melodies with lyrics on a piano roll. There is a beatbox sequencer in which every drum is a mouth gesture. It also models a songbird, a seal, an alligator and an elephant.

Jameson Hodge created the repository on November 5, and it had about 1,870 stars and 224 forks by November 11, per GitHub. It is MIT-licensed and free, with a web version and a command-line tool that renders WAV files. The architecture notes say the code was written with AI coding agents. That is the only place AI enters the project.

The method itself is old. The notes credit work going back to James Flanagan's speech research and Dennis Klatt's 1979 duration rules, and tune vowels against a 1995 study of American English vowels. What is new is a polished, free app built on top of it.

That matters for anyone using synthetic voices commercially. A physical model holds no recordings of anyone, so there is no person's voice inside it to clone, and none of the consent and licensing fights around voice cloning apply. The trade-off is control: you shape the voice by moving a simulated mouth.

Not every convincing synthetic voice is AI. For creators who want a voice with no training data behind it, CHIMERA is free to try today, and its license allows commercial use.