Attribution
on percentage-level accuracy
SUREEL is carefully tuned to best match human perception, while being based on the hard mathematical facts that determine the creation of each artwork. Thus, we guarantee that the influence of each artist is recognized and compensated properly.
Attribution Explained
Measure the influence of human creators on AI art.
We observe the influence of each artist on the AI model during training. In addition, we analyze how each new prompt is processed. Combined, we can precisely calculate the influence each artist has on AI-generated content.
With Sureel's attribution, every artist who contributes to a newly generated image is compensated. Users can mix and match different human styles to create new art, while recognizing and honoring the original creators.
Our Products
We offer attribution in two forms
Decoupled AttributionSureel Engine only requires the input, intermediate output, and final output of an AI model. The AI model can be either supplied by the customer, or the Sureel API can be used, which offers a wide range of AI capabilities.
Integrated AttributionIn our integrated option, we offer the most precise and detailed attribution possible. We require all training images, the entire training history including all weights, gradients, and loss values, and the output of all encoders and other model components used for training the AI model.
SUREEL attribution works will all cutting edge AI functionalities
Text to ImageUnleash your imagination and watch your words come to life with our cutting-edge text to image feature.
Style TransferTransform your images into unique masterpieces with our innovative style transfer mode.
In-PaintingErase imperfections and restore the beauty of your photos with our advanced in-painting technology.
SUREEL ENGINE
Decoupled Attribution
Our Plug & Play Solution
In our decoupled option, the Sureel Engine only requires the input, intermediate output, and final output of an AI model. The AI model can be either supplied by the customer, or the Sureel API can be used, which offers a wide range of AI capabilities.
For setting up our attribution model, we require the data set used to train or fine-tune the customer or Sureel AI model. After encoding these images with the chosen AI model, our setup can be completed without further input.
To obtain attribution for an image during inference, only the text prompt and the generated image are required. Leveraging the encoded input, output, and information about the training set, we obtain a percentage-wise attribution to all creators from the data set used to train or fine-tune the AI model.
Patent-pending Technology for Trusted AI
With our suite of patent-pending technologies we are at the frontier of instant attribution and decentralized provenance for generative AI done right.
SUREEL ENGINE
Integrated Attribution
Full Control, Full Integration
In our integrated option, we offer the most precise and detailed attribution possible. While the decoupled attribution already offers percentage-wise attribution to individual creators, the integrated option allows tracking every single piece of artwork in the entire process.
For setup, we require all training images, the entire training history including all weights, gradients, and loss values, and the output of all encoders and other model components used for training the AI model.
During inference, the Sureel Engine requires the input and output of the AI model as well as every activation of every layer. Since detailed information about the model is required, a more comprehensive integration into the customer model is required. Alternatively, the AI model can be selected from the Sureel AI API, where integration is done entirely on the Sureel side.
© Sureel 2024