A new AI image-generation model designed to facilitate high-quality and private image creation has been developed by researchers at the University of Surrey’s Institute for People-Centred AI in partnership with the AI company Stability AI. Named **Stable Diffusion 3.5 Flash** (SD3.5-Flash), this innovative technology allows for fast, unlimited text-to-image generation directly on consumer devices, eliminating the need for cloud services.
In contrast to traditional AI diffusion models, which typically require 30 to 50 processing steps to generate images, SD3.5-Flash achieves this in just four steps. This efficiency not only speeds up the image creation process but also makes the model lightweight enough to operate on mobile phones, tablets, and laptops, significantly reducing the energy and water consumption associated with cloud-dependent AI applications.
Lenovo has secured a license for the SD3.5-Flash model from Stability AI, integrating it into its forthcoming Personal Ambient Intelligence platform, **Qira**. The model was developed by **Hmrishav Bandyopadhyay**, a doctoral researcher at Surrey, during an internship at Stability AI, with pivotal contributions from the **SketchX Lab** at Surrey’s Institute for People-Centred AI.
“Our SD3.5-Flash model allows users to create images from text descriptions entirely on their device, with no data leaving their hardware,” Bandyopadhyay stated. “Achieving this level of efficiency is technically challenging, as it requires compressing a diffusion model to run in only a few steps while maintaining quality. It’s a great example of how academic research and industry collaboration can push the field forward, and it’s incredibly rewarding to know the technology will be used by millions of people worldwide.”
The development builds on previous research within the SketchX Lab, demonstrating how large diffusion models can be compressed into efficient versions without sacrificing image quality. This initiative is part of a broader research agenda to make generative AI more practical for real-world applications, following earlier projects such as **NitroFusion**, which focused on high-fidelity single-step diffusion.
Professor **Yi-Zhe Song**, Director of the SketchX Lab and Co-Director of the Institute for People-Centred AI, emphasized the importance of local AI solutions, stating, “At Surrey, we believe AI shouldn’t be limited to data centres and cloud infrastructure; it should run locally on the devices people use every day. SD3.5-Flash puts a powerful creative tool directly in users’ hands while keeping their data private and reducing the energy demands associated with cloud processing. For AI to be genuinely useful and responsible, it needs to be local, efficient, and accessible to everyone.”
Professor **Adrian Hilton**, Founder and Director of the Surrey Institute for People-Centred AI, echoed these sentiments, noting, “The model reflects the University of Surrey’s people-centred approach to AI, which focuses on designing technologies that are trustworthy, inclusive, and easy for people to use in their everyday lives. This approach helps ensure that advanced AI tools are not only powerful but also transparent, energy-efficient, and accessible to a much wider range of people, supporting greater trust and confidence in how AI is deployed.”
This collaboration also underscores the opportunities for Surrey students to gain industry experience while studying, involving them in projects that require both technical and professional skills for their future careers. As AI technology continues to evolve, initiatives like SD3.5-Flash illustrate a commitment to making advanced generative systems more efficient, practical, and user-friendly.
See also
Sam Altman Praises ChatGPT for Improved Em Dash Handling
AI Country Song Fails to Top Billboard Chart Amid Viral Buzz
GPT-5.1 and Claude 4.5 Sonnet Personality Showdown: A Comprehensive Test
Rethink Your Presentations with OnlyOffice: A Free PowerPoint Alternative
OpenAI Enhances ChatGPT with Em-Dash Personalization Feature
