[ad_1]
Every professional, student, and more is taking to AI tools with gusto. And why not? After all, getting AI to serve you as your personal assistant is just what the doctor ordered to lower stress levels and extended hours spent bent over a desk. AI tools are definitely a blessing for all those who have jobs to do and assignments to submit. So, if you are looking for an AI tool too that can generate high-resolution images, without the requirement of a hefty investment and resources, then we have your back. You can look at some of these tools provided by centralised services, but access depends on payment. But not everyone can, or wants, to pay for these services. Breaking free from this problem, a groundbreaking AI tool developed by the University of Surrey is set to transform the narrative, making high-resolution image generation accessible to a broader audience. Meet, DemoFusion.
DemoFusion is a pioneering model designed to democratise cutting-edge technology and that means you can access its services without the need for costly subscriptions or powerhouse computers. DemoFusion, now publicly available, aims to enrich lives by providing universal access to advanced AI image generation. In fact, DemoFusion operates seamlessly on simple consumer-grade devices. .
DemoFusion’s User-Friendly Interface
Essentially a plug-and-play extension of the open-source Stable Diffusion XL (SDXL) model, DemoFusion takes image resolution to new heights, offering a 4x, 16x, or even higher without the need for extensive coding or additional training. The only compromise, as acknowledged by the development team, is a bit of added patience, about six minutes, as tested by TNW.
From Low-Res to Brilliance
The innovative approach involves generating low-resolution images initially, followed by a progressive upscaling process to enhance detail and resolution using the SDXL model. Professor Yi-Zhe Song, the mind behind this breakthrough, highlighted the significance of the technique for the TheNextWeb by saying, “For the first time, our unique technique lets users enhance their AI-generated images without the need for vast computing power or any re-training of the model.”
Professor Song further emphasised the democratisation of digital art and imagery, envisioning a world where everyone, not just wealthy corporations, can harness this powerful medium.
Whether DemoFusion will stand shoulder to shoulder with giants like OpenAI’s DALL-E remains to be seen. Nonetheless, its creation marks a pivotal stride towards unlocking the full potential of AI-driven image generation for the public and the broader tech community. The online availability of the technique is outlined in the paper titled “DemoFusion: Democratising High-Resolution Image Generation with No $$$.”
How it is done
In its abstract, DemoFusion reveals, “Our novel DemoFusion framework seamlessly extends open-source GenAI models, employing Progressive Upscaling, Skip Residual, and Dilated Sampling mechanisms to achieve higher-resolution image generation. The progressive nature of DemoFusion requires more passes, but the intermediate results can serve as “previews”, facilitating rapid prompt iteration.”
[ad_2]
Source link