Autolume

A Neural-network based visual synthesizer.

Autolume is a no-coding generative AI system allowing artists to train, craft, and explore their own models.

Autolume Main Features

Network Training

Autolume provides the option to train models from scratch or resume training from a previously trained model. The training module accommodates both square and non-square datasets and leverages augmentation algorithms to facilitate training with smaller datasets.

Latent Projection

The projection module allows loading a target image, a text prompt, or a combination of both, to find the closest corresponding latent vector of a trained model. Different methods are provided to search for the corresponding latent vector that you can choose from.

Model Mixing

This module provides the ability to mix two trained models together to make a new model. It works based on selecting parts of one model and parts of a different model, thereby mixing the features of the two models. Model Mixing is available both offline and in the real-time module.

Feature Extraction

This module identifies and extracts interpretable feature directions in the latent space of a trained model based on Principal Component Analysis (PCA). These extracted feature directions can be then accessible in the Renderer.

Real-time Generation

Autolume-live is the module in which you can load a model and work with it in real-time. It provides you with a diverse set of features to explore the latent space in different ways and to play with the parameters of the network. It also enables you to control the parameters via OSC protocol for audio-reactive and other kinds of interactive works.

Super-resolution

The Super-resolution module provides the feature to upscale images and videos for higher resolution distribution of your work.

Where the real-time AI generation happens…

Autolume-live is the module in which you can load a model and work with it in real-time. It provides you with a diverse set of features to explore the latent space in different ways and to play with the parameters of the network. It also enables you to control the parameters via OSC protocol for audio-reactive and other kinds of interactive works.

A complete overview of Autolume features:

Autolume Technical Documentation

Autolume Artworks

Autolume Mzton

Autolume Mzton

Autolume Acedia

Autolume Acedia

Dreamscape

Dreamscape

Reprising Elements

Reprising Elements

Install Autolume

Please fill out this form to get access to Autolume installation guidelines.

To successfully install and access all Autolume features, you need a Windows or Linux system with an NVIDIA GPU (RTX1080 and above).

Report a Problem

As a dedicated research team behind Autolume, we emphasize that our system is a tool for exploration and innovation without direct support. However, we greatly value feedback from our users.

Reporting any issues you encounter with Autolume will help us with the continual improvement and refinement of our system.

Autolume Resources

Pre-trained Models

Access and use a curated list of pre-trained models for Autolume.

Datasets

Find a curated list of open-access datasets that you can use for network training.

Code Repository

Contribute to the development of Autolume.

Share Your Work

Share your datasets, trained networks, and artworks with a larger community of artists.