A free, open-source physical-modeling voice synthesizer that runs in the browser — every sound is generated by simulating air moving through a throat, mouth, and nose, not by neural models or recorded samples.
Last verified · 2026-11-09 · by Moe Ameen
Chimera Vocal Synth (shown in the app simply as CHIMERA) is a physical-modeling voice synthesizer that runs entirely in the browser. It is open source under the MIT license, free to use, and built by the GitHub developer jamesondh. The idea that defines it is on its own front page: "a voice grown from a simulated airway." Instead of playing back recorded voice samples or generating audio from a neural model trained on real singers, Chimera produces every sound by simulating the physics of a human vocal tract — lungs pushing air, vocal folds (or, for creatures, a syrinx) vibrating, and that pressure wave shaped by a tongue, lips, jaw, and nasal port. That makes it a genuinely different animal from the neural voice tools most creators know, like [ElevenLabs](/ai-tools/elevenlabs) or [Suno](/ai-tools/suno).
Under the hood the engine is written in Rust and compiled to WebAssembly, where it runs in an AudioWorklet so the simulation happens in real time as you play; a native command-line build exists for offline renders. The interface is split into a few modes. You can **play** by dragging the tongue, lips, and jaw inside a live head cutaway, using a vowel pad and a computer or Web MIDI keyboard. You can **speak** by typing English text or raw IPA, which it converts into articulatory gestures and prosody. You can **sing** on a piano roll with lyrics written one syllable per note. There is a **beatbox** mode where every drum is a mouth sound — "lips, tongue and larynx pushing, sealing and sucking air" — driven by pads and a step sequencer. And there is a bestiary of **creature** voices, each modeled from its own sound source and airway: a two-voice songbird syrinx, a seal, an alligator, and an elephant. The body itself blends continuously between male, female, and child.
What Chimera is not is a content pipeline. It is an instrument and a research demo — an exceptional one — but its output is raw browser audio. It does not write captions, cut video, wrap your sound in a visual, generate a carousel or a blog from the same idea, keep a consistent brand voice across a week of posts, or publish anywhere. You make a sound; getting that sound in front of an audience is a separate job entirely.
Chimera's whole appeal is that it makes sounds nothing else does — a synth that is pure vocal-tract physics will give you a sung hook, a beatboxed groove, or an alligator call that no stock library and no neural voice can reproduce. That is a gift for a scroll-stopping post. The catch is that it stops at the sound. Chimera has no video, no captions, no visual, and no publish button, so a one-of-a-kind clip just sits in a browser tab. [Kompozy](/) is the layer that turns that raw audio into something a feed will actually watch and spread.
Here is the concrete handoff. Record Chimera's output, then bring the audio into Kompozy as the track under a short: a [Listicle](/glossary/clipped-short) or Naturalistic video lays your bullet cards over a portrait clip while your Chimera sound carries the audio, or a Marketing Short uses the weird vocal riff as its 4-second hook before the demo footage lands. Kompozy burns on-brand, word-synced captions over the top — essential, because an unusual synthesized voice is exactly the kind of audio viewers want to read along with. Then it does the part Chimera can't touch at all: from the same idea it generates the formats that surround a drop — brand-exact [Carousel Posts](/glossary/hyperframes) and quote cards through HyperFrames, a short blog explaining how the sound was made, an email to your list, and an avatar-narrated [Persona Short](/glossary/persona-shorts). Everything is held to one [Persona Brief](/glossary/persona-brief) so the whole release sounds like one creator, and [Autopilot](/glossary/autopilot) schedules and publishes the set across the eight social platforms plus blog and email from a single queue, behind a per-post review gate. Chimera invents the sound; Kompozy ships the campaign around it.
Chimera Vocal Synth (CHIMERA) is a free, open-source voice synthesizer that runs in the browser. It is a physical-modeling synth: every sound is generated by simulating air moving through a vocal tract — lungs, vocal folds, tongue, lips, and nose — rather than by playing samples or running a neural model. It can sing, speak, beatbox, and make modeled creature calls.
Yes. The project is released under the MIT license and is free to use, built by the GitHub developer jamesondh. The engine is written in Rust and compiled to WebAssembly, with a browser app and a native command-line build for offline renders.
Not in the usual sense. ElevenLabs and Suno generate audio with neural models trained on real recordings. Chimera uses no training data and no samples at all — it is a physics simulation of a vocal tract, so its character comes from modeling airflow and anatomy rather than from a learned model. The result is a distinct, sometimes deliberately inhuman sound.
Chimera makes the audio, but it has no video, captions, or publishing of its own, so you need a second tool to turn a clip into a post. Record the sound, then use a content engine like Kompozy to wrap it in a captioned short, generate surrounding formats, and schedule it across platforms.