Frequency modulation
One signal shapes another. Sound, time, depth, pointer, physics, and speech can drive the same property system.
FM 112.5carrier / modulator
FM creative system
dreams.fm
Frequency becomes form.
Direct picture, type, sound, code, and interaction from one live scene graph. Work by hand, work with the model, or move between both without flattening the result.
Foundation model
One model understands the structure. It writes validated scene patches instead of replacing the work with an opaque result.
Everything is a scene
One composition. Every surface.
Picture, type, sound, code, and interaction remain in the same project. The canvas, timeline, model, and runtime all address the same scene graph.
01
Input
Begin with a prompt, script, media, composition code, or an empty scene.
02
Structure
Intent resolves into typed scenes, artifacts, layers, and exposed properties.
03
Modulate
Route time, sound, interaction, depth, and logic into any addressable property.
04
Transmit
Render video, publish a site, embed a live scene, or keep the work responsive.
Same instrument
Different language. Same signal.
The visual language can change completely. The contract does not: every artifact stays addressable, routable, and editable.
Foundation model
The model edits the instrument.
Work directly, import code, or direct the model in plain language. Every route produces the same validated mutations, history, selection, and undo.
Read the LiveEngine contractScene graph
Every visible object remains selectable, addressable, and editable on canvas.
Shared clock
Scenes, transitions, speech, music, modulation, and interaction stay deterministic.
Spatial field
Compose 2D, true 3D, depth maps, parallax planes, materials, and camera paths.
Signal routing
Patch waves, envelopes, beats, bands, speech, and live audio into exposed properties.
Token economy
Small patches. Deep scenes.
Scenes reference reusable artifacts and expose meaningful properties. The model changes compact structured data instead of rewriting the production on every turn.
scene.dreams.jsonvalidated
{
"scene": "launch-reveal",
"artifacts": ["depth-city", "kinetic-title"],
"signals": {
"bass.attack": "city.cars.stretch",
"voice.energy": "title.bloom"
},
"transition": "beat-cut"
}Private beta
Help tune the first public release.
We are inviting directors, designers, developers, educators, and experimental storytellers into the Studio in small groups.
- One email when your access is ready
- No credit card and no marketing sequence
- Your answer helps us prioritize the right workflows
Access
Before you join.
What can I create with Dreams.fm?
Dreams.fm is built for multi-scene video, motion graphics, websites, presentations, visual documents, embeds, live assistants, and interactive worlds. These formats share one scene-native project instead of becoming disconnected exports.
Do I have to use the model?
No. The visual canvas, timeline, animation tools, 3D controls, audio routing, artifact library, and code composition workflow can all be used manually. The model is another way to operate the same Studio, not a requirement.
What remains editable after generation or import?
Native projects retain addressable scenes, layers, text, media, transforms, animation, signals, effects, transitions, and exposed artifact properties. Imported composition code is standardized into that project model while its source remains available for download and reuse.
How does private beta access work?
Join with an email and tell us what you want to make first. We invite small groups as the relevant workflows are ready and send one email when access is available. No credit card is required to join the list.