AI Synth Plugin: Recreate Any Sound with Chatmu Sound

Chatmu Sound is an AI synth plugin that programs itself: you describe a sound in plain words, or hand it a recording, and the Chatmu agent builds the patch for you with pure synthesis. It is a sound-design synth based on the open-source Vital engine, it runs inside Ableton Live through Chatmu Desktop, and every knob it turns stays editable.

Chatmu is an AI agent for the music industry with its own catalog of music-industry MCP connectors. Chatmu Sound is one of the four plugins that ship with Chatmu Desktop on October 31, 2026.

Updated October 2026.

What is Chatmu Sound?

Chatmu Sound is a wavetable synth that the agent can play like an instrument and program like a sound designer. Under the hood it is a fork of Vital, the open-source wavetable synth created by Matt Tytel. Vital is one of the best synths ever given away to producers, and we built on it because its engine is deep enough to cover basses, pads, leads, plucks and realistic tones. Like Vital, Chatmu Sound is licensed under the GPLv3; the original project lives at github.com/mtytel/vital. Credit for the synthesis engine belongs to Matt and the Vital community. What we added is the layer that lets an agent talk to it.

That layer is a local control channel: every parameter in the synth (oscillators, wavetables, filters, envelopes, LFOs, effects) is exposed to the agent. So when you ask for something, the agent is not picking a preset from a list. It is setting the actual parameters, the same ones you would touch with the mouse.

I'm a producer and sound engineer, and I've spent more hours than I'd like to admit scrolling through preset browsers looking for "that sound". Chatmu Sound is the tool I wanted: say what you hear in your head, get a starting patch in seconds, then make it yours.

How does an AI synth plugin program a patch from a description?

The first workflow is text. You type (or say, with push-to-talk) what you want, and the agent programs the patch on the Chatmu Sound instance in your Live set. A few prompts that work well:

"Warm 808 with a short tail, a bit of saturation, tuned to F." The agent sets a sine-based oscillator with a pitch envelope for the punch, shortens the amp release, adds drive and tunes it to the key of your song.
"Wide supersaw pad for the chorus, slow attack, darker than the lead." It stacks detuned saw voices, spreads them in stereo, opens the attack and pulls the filter down so the pad sits under the lead instead of fighting it.
"Growl bass for a dubstep drop, moving at 1/8." It builds a modulated wavetable bass with an LFO synced to the tempo of your session, and a filter movement that gives the "talking" quality.

Because the agent works inside Chatmu Desktop, it knows your session: the tempo, the key from your MIDI clips (Chatmu reads the chord on each bar), and what else is playing. "Make the pad darker" or "shorten the tail of the 808" is a follow-up, not a new search. The patch is a normal synth patch, so you can open the interface and keep tweaking by hand at any point.

Can AI recreate a sound from a recording?

Yes, and this is the workflow I'm most excited about. The second way to use Chatmu Sound is to give it audio: a trumpet line, a guitar note, a plucked string, a bass from a demo. The agent analyses the recording and recreates its timbre with pure synthesis, then loads that patch into Chatmu Sound.

It is important to be precise about what this is. It is not sampling. Chatmu Sound does not play back the recording or slice it into a sampler. It rebuilds the sound from oscillators, filters and envelopes, so what you get is a synth patch that sounds close to the reference. The audio is a target, not a source.

Recreating the patch is half of the job. A trumpet does not sound like a trumpet only because of its tone; it also bends, swells and breathes. So along with the patch, the agent writes playable MIDI with the expression it detected in the reference, and sends it into Live as a clip. You get a track you can play, re-voice or rewrite, not a frozen audio file.

A typical request looks like this:

"Here's a 2-bar trumpet from my old demo. Recreate it in Chatmu Sound and put the melody on a new MIDI track at 92 BPM."

The agent creates the track, loads Chatmu Sound, programs the patch, writes the clip and tells you what it changed. If the result is close but too bright, you say so and it adjusts.

Why synthesis beats samples for producers

Samples are great, and Chatmu can even search Splice for you. But for a sound you want to own and shape, a synth patch has real advantages:

Fully editable. Every part of the sound is a parameter. Change the attack, the brightness, the vibrato depth or the length of the tail without re-recording anything.
Playable in any key and tempo. No time-stretch artefacts, no transposition that turns a cello into a chipmunk. It is an instrument, so it plays whatever notes you write.
No licensing grey zone. Because the sound is rebuilt with synthesis rather than copied from the recording, you are not lifting someone else's audio into your track. You still need to respect the rights of any melody or composition you reuse; the timbre is what gets rebuilt.
Small and portable. A patch is a small preset file, not gigabytes of multisamples. It travels with your project.
A starting point, not an endpoint. A recreated trumpet can become something no trumpet can do: add a wavetable morph, a filter sweep, a granular-feeling tail. That's sound design.

This fits the core idea behind Chatmu Desktop: it's not music made by AI. It's AI inside YOUR DAW. You decide, everything stays editable, the stems are yours, and it sounds like you.

How does the agent know the sound is right?

Most AI sound design tools stop at "here's a patch, good luck". Chatmu has ears. Chatmu Ears is the listening engine inside Chatmu Desktop; it measures audio in real units: spectrum, loudness, dynamics, stereo image and loudness-matched A/B comparisons.

When you recreate a sound from a recording, the agent renders the patch, compares the result against the reference with Ears, and keeps adjusting parameters to get closer. It looks at things a producer would check by ear: is the spectral balance right, does the attack hit the same way, does the sound decay at the same speed. When it reports back, it can tell you where the match is close and where it isn't ("the body matches, the breath noise on the attack is weaker than the reference").

The same ears work on your whole mix. With the Coproducer plugin on Ableton's Master, Chatmu listens to your mix when you hit play, so you can ask "does the new bass mask the kick?" right after programming it.

AI sound design tools in 2026: how does Chatmu Sound compare?

Chatmu Sound is not the first tool to connect AI and synthesis, and it would be unfair to pretend otherwise. Here are the ones worth knowing, based on their public pages:

Synplant 2 by Sonic Charge introduced Genopatch, which analyses a short audio sample and grows synth patches that get closer to it, all running locally on your computer. It's a respected, creative take on audio-to-patch, built around Synplant's own synth engine.
Fadr SynthGPT is a plugin that lets you create instruments from a text description and play them in your DAW; according to Fadr's docs, the prompt is processed by Fadr's AI engine.
Tilichu by Butterlamp Audio is a wavetable synth launched in August 2026 with a built-in generator that turns a text description into a full preset (Sonicstate coverage).

Tool	Input	Output	Part of a wider agent
Chatmu Sound	Text or a recording	Editable Vital-based patch + playable MIDI in Live	Yes, the Chatmu agent in Chatmu Desktop
Synplant 2 (Genopatch)	Audio sample	Editable Synplant patch	No, standalone instrument
Fadr SynthGPT	Text	Playable instrument in the plugin	Part of Fadr's suite of music tools
Tilichu	Text	Full preset in its wavetable synth	No, standalone instrument

The difference with Chatmu Sound is not a single feature; it's context. The synth is one tool in an agent that also hears your mix, writes MIDI, arranges tracks in Live, and later takes the exported song to distribution, video and social. If all you need is a patch generator, any of the tools above can be the right pick. If you want the patch to land on a track, in key, with the melody already written, that's what we built.

Where is Chatmu Sound available?

Chatmu Sound ships with Chatmu Desktop, launching Saturday, October 31, 2026, for Mac (Apple silicon and Intel) and Windows. Version 1 works with Ableton Live 11 and 12, starting with Ableton; more DAWs are coming. It installs alongside the other three launch plugins: Chatmu Coproducer (the bridge on the Master so Chatmu hears your mix), Chatmu Instrument (the agent designs a playable synth for you, interface included) and Chatmu FX.

Reading your session is free; for the agent to create tracks or load plugins you turn on "Allow control", and destructive actions always ask first. There's a free plan, and you can see the details on the pricing page. If you already use Chatmu in Claude or ChatGPT through our Chatmu MCP, the same account works. For the bigger picture of how MCP connects AI to music tools, read our complete guide to MCP for the music industry.

FAQ

Is Chatmu Sound an AI synth plugin or a sampler?

It's a synth. It recreates sounds with pure synthesis (oscillators, filters, envelopes), so the result is an editable patch, not a playback of the original audio.

Is Chatmu Sound based on Vital?

Yes. It's a fork of Vital, the open-source wavetable synth by Matt Tytel, and it's licensed under the GPLv3 like the original.

What kinds of sounds can it recreate?

Synth sounds such as basses, pads, leads and 808s, and acoustic sources like trumpet, guitar, plucked strings or bass. Results are closest when the reference is a clean, isolated note or phrase.

Does it make music for me?

No. It's not AI-generated music. The agent programs sounds and writes editable MIDI inside your session; you decide what stays.

Which DAWs and systems does it support?

Chatmu Desktop runs on Mac and Windows and works with Ableton Live 11 and 12 at launch. More DAWs are coming.

AI for the industry. Humans for the music.

Try Chatmu → | See pricing →

]]>