genereti Introduction
genereti is an open-source collection of more than 100 nodes that brings live visuals, sound, drawing, and media into ComfyUI. Instead of using a workflow only to produce a finished image, you can use it as a live creative instrument: edit a sketch, play a video, change a sound patch, or draw a stroke and see the connected parts respond in your browser.
Genereti is especially useful when you want to experiment with motion and interaction. You might draw a shape and send it through a chain of visual effects, build a music video whose elements respond to shared timing, or use movement from a film to control another part of a workflow. More than 40 example workflows provide starting points you can explore and change.
A Genereti ComfyUI graph with audio cables that move with the sound
Watch the one-minute demo to see a workflow in action, or visit the project page for an overview.
Early alpha: Genereti was built for classroom use and live performance. Bugs and interface changes are possible, so back up workflows you want to keep before updating.
How genereti Works
Think of a ComfyUI workflow as a set of creative tools connected by wires. Each node does a job, and its output can become another node’s input. In Genereti, those jobs can happen live: a sketch can make an image, an effect can transform it, and a final node can show the result as you work.
For example, you could draw in an Excalidraw node, connect that drawing to a feedback effect, and watch each new stroke change the image downstream. Or you could make a visual in p5, then connect its output to a second sketch that turns it into mirrored tiles. You can explore these ideas by changing one node at a time and watching how the connected result changes.
Genereti also connects different kinds of activity. Its modular audio nodes let you build sound from parts such as sequencers, synths, filters, delays, and a mixer. Its optional perception nodes can identify pose, hand, or face information in a film or camera feed, so movement becomes data you can connect elsewhere in the workflow.
The standard demos run in the browser and do not require a model download, a separate model server, or another node pack. You need a recent ComfyUI and a browser with WebGPU support, such as a recent Chrome or Edge. The optional perception workflow downloads its browser model when you request it.
genereti Features
Live visuals and code
Genereti lets you edit visual code inside a node and see it run live. Supported approaches include p5, GLSL shaders, Three.js, tixy, HTML, SVG, and LaTeX. These are different ways to describe visuals: for instance, a p5 sketch can generate an animated image, while a shader can change how an image is rendered. You do not need to learn them all to begin; the example workflows let you edit an existing result and observe what changes.
A Three.js torus knot beside its code
You can also chain sketches. One sketch’s image output can feed another sketch or effect, just as an image can pass between other ComfyUI nodes. This makes it possible to build a look in stages rather than trying to create everything in one place.
A p5 source feeding mirrored tiles
Modular audio and performance timing
Genereti’s browser-based audio tools include a transport, sequencers, a drum kit, synths, filters, delays, and a mixer. You can connect these parts to build a sound patch and adjust its components while you listen. To start sound in an example workflow, press ▷ on a mod.output node.
A performance timeline provides shared musical time for a workflow. This helps when you want visual and sound elements to follow the same performance. Press Alt+Shift+T to open the timeline in the ꘇ-Performance-Timeline example.
Synth envelope and filter controls
Drawing, media, and texture effects
Excalidraw and Poliplan are available as drawing nodes. A drawing can serve as the beginning of a larger workflow: connect its output to effects and see new strokes influence what follows.
An Excalidraw spiral driving a feedback chain
You can also work with videos, GIFs, and images. Texture operators include feedback, colormaps, bloom, and displacement. Try connecting a source to different operators, or changing their order, to explore how the image develops.
Film footage processed through a feedback chain
Optional movement and face tracking
Genereti’s optional MediaPipe perception tools can detect pose, hands, and face information from a film or camera. A region picker lets you choose areas such as wrists, torso, lips, or eyes. You can then use the selected movement as input for drawing, sound, or events elsewhere in the graph.
The ꘇ-Perception-01-Streams example uses MediaPipe on a film and downloads its browser model when requested. You do not need perception features to use Genereti’s standard demos.
Picker for body, face, and hand regions
Other workflow tools
Genereti also includes webview pages for notes and math, image comparison views, and an optional assistant. The node catalog is a useful place to explore what is available beyond the first examples.
genereti Models
Genereti itself does not need an image-generation model download for its standard demos. Two model-related options are worth distinguishing:
- MediaPipe browser model: Used by the optional perception workflow to identify pose, hand, or face information in a film or camera feed. It downloads on request, so you only need it when trying those features.
- GeneretiCore: A separate, experimental project, not a requirement for Genereti. It uses local models on Apple silicon to respond to drawing live. See the GeneretiCore repository if that particular drawing-to-model experience interests you.
Drawing into a local ink-wash model with the separate GeneretiCore project
Troubleshooting genereti
Because Genereti is an early alpha, back up an important workflow before updating or making substantial changes. For a first test, start with an example workflow; this gives you a known starting point before you build your own graph.
I cannot find the example workflows. What should I check?
- Restart ComfyUI and refresh its browser page.
- Look under Workflows → Genereti or Templates → genereti.
- If they still do not appear, consult the student quickstart and documentation index.
The audio patch is visible, but I cannot hear it. How do I start it?
- Find the
mod.outputnode in the workflow. - Press its ▷ button to start sound.
- For a small patch to explore, open ꘇ-Modular-Audio.
A standard demo asks me to download a model. Is that expected?
The standard demos do not need model downloads. The optional ꘇ-Perception-01-Streams workflow is different: it downloads a MediaPipe browser model when you request its perception features. GeneretiCore is a separate project and is not needed for Genereti.
Which browser should I try?
Use a browser with WebGPU support, such as a recent version of Chrome or Edge, alongside a recent ComfyUI.
I want to understand a workflow without starting from scratch. Where should I begin?
Open ꘇ-Performance-Timeline for a tour of visuals, sound, and timing; ꘇ-Livecode-Languages to compare six livecode approaches; or ꘇ-Livecode-P5-Pipeline to see one sketch feed another. Change one part at a time so you can see what that node contributes.
Learn More about genereti
- Student quickstart: a gentle starting point for exploring Genereti.
- Livecode tutorials: guided help for making and changing live visuals.
- Full workflow list: browse the example workflows by what you want to try.
- Documentation index: find guides by area.
- Node catalog: look up the available nodes.
- Modular audio guide: learn more about building sound patches.
- Performance timeline guide: explore shared timing for live work.
- Keyboard shortcuts: find controls for working in the interface.
- Project page and one-minute video demo: see Genereti in action.
