The first and only
AI-native instrument
Play, improvise, and explore with live audio synthesis for music and sound design.
Music doesn't fit in a prompt.
Music and sound design are spontaneous, delightful, and alive. We deserve tools that harness the best parts of AI and cut the rest.
Daydream is the first AI-native instrument,
with every detail designed for human control
~81ms end-to-endYour hand
step 1
MIDI · knobs · code
Instrument
step 2
Web · VST · TD · API
DEMON
step 3
Open-source engine
Audio
step 4
48kHz · 25Hz steerable
Inside DEMON · latent navigator
Five ways to play.
However you create, Daydream meets you there. Open it in a browser, load it in your DAW, drive it from TouchDesigner, call it from code, or run the whole engine on your own GPU.
Heard from artists.
“DEMON is not just another prompt-to-song system. The real-time control layer changes the shape of the problem. Being able to move a parameter and hear the music respond almost immediately makes the model feel closer to an instrument, a synth, or a live performance system than an offline generator.”
Built for the
joy of creation.
Daydream gives you unprecedented control to sculpt tracks and soundscapes — whether you're producing music, performing a set, or scoring a film.
MIDI-MAPPING
Map the model's latent space directly to physical hardware faders, knobs, and macro lanes. Because control propagates at 25Hz, the audio stream reshapes the exact millisecond your hand moves a controller.
LIVE REMIXING
Load a track and bend it on the fly. DEMON streams variations continuously with sub-100ms latency, allowing you to re-style and morph arrangements live without breaking your performance flow.
STYLE EXPLORATION
Audition entirely new genre trajectories, continuous key shifts, and unexpected arrangement variations on your timeline without waiting for a render.
SEPARATE STEM CONTROL
Manipulate tracks without destroying the core hook. Isolate and lock vocal elements or primary melodies while continuously diffusing and mutating the underlying instrumental stems beneath them.
We build in the open.
Daydream's engines are open-source, free to run locally and developed in public.
Real-time audio generation engine
Diffusion-native engine for continuous frame-level audio control. Sub-100ms parameter latency on a single consumer GPU. Powers every Daydream instrument.
Real-time AI video generation
Point any source at it, pick a model, and watch generation happen in real-time. On your machine or in the cloud.
Join the Community.
Hear What's Next.
Join our Discord to get early previews of new tools, share your work, and join live jam sessions. Whether you are debugging code, producing tracks, or designing sound, this is the place to get feedback, technical support, and find collaborators.
Frequently asked questions
Suno and Udio are batch-mode cloud services — type a prompt, wait for a finished song to render, start over to change anything. Daydream takes the opposite approach: you twist a knob and hear the change in about 81 milliseconds. The model responds to your performance instead of replacing it. There's no prompt-and-wait loop — you play it like an instrument.
No. The web app at music.daydream.live runs in your browser — no install, no GPU required. Load a sample, twist parameters, and hear it remix instantly. The VST plugin and TouchDesigner operator are coming, and you can self-host the DEMON engine on your own GPU today.
Yes — Daydream ships as a native VST plugin (AU coming soon). It integrates with your project timeline so you can automate generation curves using standard envelope lanes, macro knobs, and MIDI hardware. Join the early-access list at music.daydream.live/alpha?key=DemonVSTAlpha2026.
Yes. Map the model's latent space directly to physical faders, knobs, and macro lanes. Control propagates at 25Hz, so the audio stream reshapes the exact millisecond your hand moves a controller. MIDI mapping is supported in the web app today and will be first-class in the VST plugin.
Our engine is. DEMON — the real-time audio generation engine behind every Daydream instrument — is open source and free to run locally on your own GPU. The web app and VST run on our hosted infrastructure (you can subscribe to a plan), but the underlying engine is fully open. Scope, our video project, is open source too.
Yes — Daydream exposes an MCP server and API for programmatic control over every parameter. Build agentic workflows, generative installations, custom apps, or anything else you can wire up. Join the API waitlist at tally.so/r/ja5Rex.
Not for any of our hosted products. The web app, VST plugin, and TouchDesigner operator all run on Daydream's cloud — any laptop with a modern browser or DAW will work. A GPU is only required if you want to self-host the DEMON engine locally, in which case an NVIDIA card with at least 8GB of VRAM works (RTX 4090 or 5090 are the sweet spot for sub-100ms latency).
Google's Magenta Real-Time 2 is an outstanding open-source generation model — it's complementary to DEMON and will likely be integrated as a supported model in the future. The framing difference: Magenta focuses on the model itself, we focus on usability as an instrument. At the moment Magenta also has noticeably higher parameter latency (~200ms) than DEMON, which has real implications for playability — sub-100ms is the threshold where the engine starts to feel like an instrument under your hands.
Today: ACE-Step v1.5 (2B turbo + 5B XL turbo). Stable Audio 3.0 and Magenta RT2 coming soon.