Shulou(Shulou.com)11/24 Report--

Recently, Adobe released its own AI image generation model suite Adobe Firefly, which officially entered the AIGC battlefield.

The iPhone time for AI has come.

At the GTC conference, Lao Huang Guanxuan and Adobe unveiled the artificial intelligence image generation tool Firefly (Firefly).

(address: https://firefly.adobe.com/) according to the announcement, Firefly has been integrated into Adobe's enterprise creative tool Adobe Express for internal testing, and only WordArt is available.

At present, it is not open to users, but you can apply for backup.

In the future, this tool will be added to Adobe's Photoshop, Illustrator and Premiere.

"with fireflies, everyone can create the content of their dreams in their own language. "

Fireflies fly into AIGC currently, Firefly provides two modes, one is the ability to generate images using text, and the other is font effect editing.

In drawing, Firefly is similar to the popular AI drawing tool, which can generate images according to prompt, provide hundreds of styles and adjust the results.

In addition, Firefly can also generate images based on the surrounding content, which allows users to test and improve the image effect more effectively.

In graphic design and illustration work, custom vectors, brushes, and textures can be generated from a few words or even sketches, and you can edit content created using familiar tools.

In addition to basic text-to-image generation, the Firefly model can "transfer" different styles to existing images like Prisma. Depending on the description provided by the user, it can also apply styles or textures to letters and fonts.

On the creative writing tools, people can create characters that look furry, scaly and other forms and effects, similar to the "art word" function in Office.

Firefly can also change 3D modeling, such as quickly generating model styles and materials:

The maximum resolution that Firefly can produce so far is 2000 pixels.

In addition, Firefly is still exploring other features, including generating vectors / brushes / templates for text, generating images based on your own style, specifying aspect ratios to expand images, generating images based on 3D models, and so on.

The "no original sin" model is here, Adobe says, and the work created by the Firefly model will contain metadata to prove that it was generated partly or entirely by artificial intelligence.

The main reason for this is to protect the copyright of artists.

Technically, Firefly is no different from DALL-E2, and text-to-image AI like Stable Diffusion.

These AI models learn to train existing images to generate new images from text prompts, usually from data sets that are pieced together by searching public image hosting sites.

These AI models can "learn" from text prompts to generate new images by "training" on existing images.

These existing images usually come from datasets obtained by searching public image hosting sites.

Some experts believe that in the United States, the use of public images (even copyrighted) for training models will be protected by the principle of fair use. However, this is a problem that is unlikely to be solved in a short time.

Midjourney and Stability AI face lawsuits after being accused of violating the rights of millions of artists for using their tools to capture images online.

Getty Images, a supplier of stock pictures, has reportedly taken Stability AI to court separately, accusing it of using millions of images on its website without permission to train the art generation model Stable Diffusion.

In addition to the unsolved issues of artist and platform compensation, there is a more pressing problem facing generative artificial intelligence:

It tends to use content such as copied images and text, including copyright content, which comes from the data used to train the model.

Some image hosting platforms have banned the use of AI-generated content because of concerns about legal disputes.

Experts warn that companies may be at risk if they inadvertently incorporate copyrighted content generated by generated AI tools into any product they sell.

Adobe's solution is to use only the company's royalty-free media library, Adobe Stock, and to train the Firefly model through copyright public license and public domain content.

Adobe said that in the future, users will be able to train and fine-tune the Firefly model with their own content to make their output have a specific style and design language.

Reference:

Https://www.adobe.com/sensei/generative-ai/firefly.html

This article comes from the official account of Wechat: Xin Zhiyuan (ID:AI_era)

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.