Stable Diffusion 3 Medium is openly released, a relatively small model ideal for personal use
Stability AI, the developer of the image generation AI 'Stable Diffusion,' has announced the release of a model of ' Stable Diffusion 3 Medium .'
Announcing the open release of Stable Diffusion 3 Medium, the most sophisticated image generation model — Stability AI Japan
stabilityai/stable-diffusion-3-medium Hugging Face
https://huggingface.co/stabilityai/stable-diffusion-3-medium
The Stable Diffusion 3 is a model that was announced in February 2024, and became a hot topic for its features such as the ability to seamlessly depict specified text within an image and to depict multiple subjects in high detail.
High-quality image generation AI 'Stable Diffusion 3' announced, enabling high-precision realization of 'depiction of specified characters' and 'depiction of multiple subjects', which are difficult for image generation AI - GIGAZINE
Stable Diffusion 3 Medium is a relatively small model with 2 billion parameters, making it ideal for running on personal systems and enterprise GPUs. Stability AI lists the following features of Stable Diffusion 3 Medium:
Overall quality and photorealism
It has excellent detail, color, and lighting, and enables photorealistic output and high-quality output in flexible styles. With innovations such as 16-channel VAE, it also addresses pitfalls common to other models, such as the realism of hands and faces.
Understanding prompts
Understands long, complex prompts that include spatial reasoning, constructs, actions, and style. You can use all three text encoders, or any combination of them, to trade off performance and efficiency.
・Text generation
The Diffusion Transformer architecture reduces spelling, kerning, typography and spacing errors for unprecedented text quality.
Resource efficiency
The low VRAM footprint allows it to run on standard consumer GPUs without performance degradation.
Fine tuning
It can understand subtle details from small data sets and is perfect for customization.
Stability AI also announced collaboration with NVIDIA and AMD. By utilizing NVIDIA RTX GPUs and TensorRT, the performance of all Stable Diffusion models, including Stable Diffusion 3 Medium, has been enhanced. In particular, the TensorRT-optimized version claims to achieve 50% performance improvement compared to the previous version.
Stability AI also announced that it is optimizing Stable Diffusion 3 Medium inference for a variety of AMD devices, including AMD APUs, consumer GPUs, and MI-300X enterprise GPUs.
The Stable Diffusion 3 Medium model data is publicly available on the online AI platform Hugging Face and is available under an open non-commercial license and a low-cost creator license at the time of writing.
In addition, at the same time as the open release of Stable Diffusion 3 Medium, the API for Stable Diffusion 3 Medium will also be available, and it can be used by the chatbot ' Stable Assistant ' and ' Stable Artisan ' running on Discord. To use either, you need to sign up for a monthly paid subscription plan, but a 3-day free trial is available.
Related Posts:
in Software, Posted by log1i_yk