StabilityAI is today releasing its new Stable Diffusion 3.5 family of AI image models. It comes with improved realism, prompt adherence and text rendering compared to SD3.
Like the first incarnation of SD3, Stable Diffusion 3.5 comes in three sizes — Large (8B), Large Turbo (8B), and Medium (2.6B). These are all customizable and tuned to run on consumer hardware, while also being available to use under the Stability AI Community License.
In short, this is a big step towards any user being able to create more realistic AI images. In a press release, StabilityAI admitted its Stable Diffusion 3 Medium model that launched in June "didn't fully meet our standards or our communities’ expectations".
The company added that "After listening to the valuable community feedback, instead of a quick fix, we took the time to further develop a version that advances our mission to transform visual media."
Our AI Editor Ryan Morrison has been testing SD3.5 and says it is a significant upgrade, matching and potentially exceeding the capabilities of the recently released Flux 1.1 Pro.
What's new with Stable Diffusion 3.5?
Stability AI says the models included are focused on customizability, efficient performance, and diverse outputs. "Stable Diffusion 3.5, our most powerful models yet, reflects our commitment to empowering creators with tools that are both widely accessible and cutting-edge," a spokesperson explained.
That means images can be fine-tuned, the models run on consumer hardware "out of the box", and the images generated will feel more unique to one another.
A quick test of SD3.5 Large found it was fast, very good at adhering to an exact prompt and stylistically controllable. It is a huge step up from SD3, especially the medium version.
Ryan Morrison, AI Editor (Tom's Guide)There's also a focus on new stylistic choices, including photography, painting, and more. It takes in hashtag prompts to specify a style such as boho or fashion. It can also use highlighting within the prompt to steer the model in a specific direction.
"Additionally, our analysis shows that Stable Diffusion 3.5 Large leads the market in prompt adherence and rivals much larger models in image quality," a press release explains.
"Stable Diffusion 3.5 Turbo offers some of the fastest inference times for its size, while remaining highly competitive in both image quality and prompt adherence, even when compared to non-distilled models of similar size."
"Stable Diffusion 3.5 Medium outperforms other medium-sized models, offering a balance of prompt adherence and image quality, making it a top choice for efficient, high-quality performance."
The model is free for non-commercial use, including scientific research, and free for small to medium-sized businesses up to $1 million in revenue. Anything beyond that will require an Enterprise Licence.
Prompt: a scene of a celestial garden floating in space, with trees made of crystal and flowers that glow in different colours. Planets and stars can be seen in the sky above, and a river of light flows through the garden, surrounded by statues of ancient gods and glowing orbs. Ethereal and serene.
More from Tom's Guide
- Apple reportedly '2 years behind' on AI with Apple Intelligence — here's why
- I write about AI for a living — here's how to become a true power user
- Watch out, OpenAI — Haiper 2.0 AI video generation model just launched and it looks stunning
English (US) ·