aura
Photo: Annie SprattABCtic
▾ var attribution = {
source_A ----[0.38]
source_B ----[0.17]
source_C ----[0.06]
model ------[ap_001]
licence --[public domain]
}- ASunset over the dunes, oil sketch0.38
- BFence and marram grass, etching0.17
- CStudy of light, watercolour0.06
Example key, for illustration
Creative AI models trained only on the public domain
Our first model, Artist’s Proof (ap_001), is under active development for release in Q1 2027. It learns only from open, free-to-use images, and every picture it makes credits its inspiration, pointing back to the works it drew on most.
Acknowledging human creation.
Current generative models are capable tools. Most were built on work used without consent, credit or payment. Auratic takes a different approach.
Taken without asking
Many of the best-known models for images, video, text and music were trained on work scraped from the web, without the permission of the people who made it, and without credit or payment.
No way to check
Creators cannot easily find out whether their work was used. Where opting out exists, it usually comes after the model has been built.
Copies and imitations
Models can reproduce near copies of work they were trained on, and can be asked to imitate a living artist, writer or musician by name.
Unchecked data
Web-scale datasets are too large to check by hand. Some image datasets have been found to contain private medical photos and illegal abuse material.
No record
Nobody can say which works shaped a given output, so the people who use these models cannot know what they are building on.
Models, tools and services
Auratic makes open models trained fairly, and the tools and services artists need to work with them.
Models
Open models trained only on works that are free to use. Artist’s Proof (ap_001), for images, is under active development for release in Q1 2027. Video models will follow.
Artist’s Proof (ap_001), release Q1 2027tools
Open source tools, built with creators, for creators, on our open models trained fairly.
Open sourceservices
Bespoke services for artists and studios, including hosting models locally on your own hardware.
Ask about services
Photo: Johannes KopfAOur first model, Artist’s Proof (ap_001)
Under active development for release in Q1 2027. It generates images from multimodal structured inputs, is trained only on works that are free to use, and every image it makes will list the sources that shaped it.
▾ var model = {
name ---------[ap_001]
training -[public domain only]
status ---[in development]
release --[Q1 2027]
sources ---[to be listed]
shown ----[every output]
}Atlas of
sources
Every image Artist’s Proof (ap_001) generates will be bound to the sources it was drawn from.
Photo: FujiphilmAB- AVilla by the sea, photograph0.41
- BWindow study, drawing0.12
Photo: A. C.AB- AHead of a man, looking up0.38
- BStudy in blue, print0.17
Photo: Annie SprattA- ACloister arcade, engraving0.44
Photographs from Unsplash, credited on each plate. The keys are examples, to show the format.
Index of plates
Each image Artist’s Proof (ap_001) generates, set beside the public domain works that shaped it.
Photo: Ivan AvilesTav. I
Photo: Eugene GolovesovTav. IIAB▾ var index = {
plates ------[on release]
sources -----[to be published]
licence ---[public domain]
model ------[ap_001]
}
▾ let tav_II = {
source_A ----[0.38]
source_B ----[0.17]
licence -----[PD]
}
Photo: Heye JensenTav. III
Photo: Rowan HeuvelTav. IV
Photo: Onur KurtTav. V
Photo: Marvin Alexander AvilaTav. VI| Tav. | Image | Largest source (example) | Share | Licence |
|---|---|---|---|---|
| I | Figure against orange light | Silhouette at dusk, albumen print | 0.41 | CC0 |
| II | Birches in autumn | Birch wood, oil on canvas | 0.38 | CC0 |
| III | Towers in motion | City blocks at night, photograph | 0.12 | PDM |
| IV | Moving water | Waves at evening, woodcut | 0.22 | PDM |
| V | Tram at night | Night tram, glass-plate negative | 0.18 | PDM |
| VI | Alley hung with buoys | Fishing village lane, etching | 0.09 | PDM |