Stability AI's new version of Stable Diffusion 3 Medium faces challenges in reproducing human anatomy, showing people with the wrong number of limbs or missing heads due to limited training data used to train the model.
Anatomical glitches
Human figure distortions
Shortly after the release, Stability AI announced its new image generation model, Stable Diffusion 3 Medium the most advanced, but limitations in the training data led to disturbances in the reproduction of the human body. The model generates frightening images of people with excess limbs, missing body parts or even without heads. The defects appear more frequently in queries with backgrounds containing grass or water, although the reasons for this are unclear.
Users speculate that the problem is due to the limited datasets used to train the model. Stability AI likely censored most of the images of people, which negatively impacted the model's ability to reproduce human anatomy. Although the stated goal was to eliminate common errors with hands and faces, the overall quality of body reproduction deteriorated.
Ambitious promises
Stability AI positioned Stable Diffusion 3 Medium as the most advanced open model for generating images from text. It was claimed that a model with 2 billion parameters would provide photorealistic results. However, in practice it demonstrates serious anatomical defects when depicting people.
Users can currently test the new model through the Stability API. Founded in 2020, the company has quickly established itself as one of the leaders in the field of generative AI, alongside competitors such as Midjourney and OpenAI's Dall-E. However, in 2023, Stability AI faced lawsuits from artists, the departure of its CEO and a cash crunch, forcing it to consider selling.
Glossary
- Stability AI is a British startup specializing in the development of generative artificial intelligence systems for creating images from text descriptions.
- Stable Diffusion 3 Medium is a new image generation model from Stability AI, billed as the most advanced open model of its type, but has encountered problems in reproducing human anatomy.
- Midjourney is a competing service for generating images from text descriptions, using artificial intelligence algorithms.
- Dall-E is a generative AI system from OpenAI for creating images from text queries.
Links
- Creepy images of people from Stable Diffusion 3 Medium on Reddit
- Test Stable Diffusion 3 Medium via the Stability API
Answers to questions
What problems are there with the new Stable Diffusion 3 Medium model from Stability AI?
Why are there problems with anatomy not being reproduced correctly in Stable Diffusion 3 Medium?
What promises did Stability AI make regarding Stable Diffusion 3 Medium?
What other problems have Stability AI been having lately?
What position did Stability AI have in the generative AI market?
Hashtags
Save a link to this article
Discussion of the topic – New Stable Diffusion 3 Medium from Stability AI: creepy mutants and strange restrictions
Stability AI's new image generation model, Stable Diffusion 3 Medium, although described as "the most advanced," has encountered problems reproducing human anatomy due to limitations in the neural network's training data, producing creepy images of mutant humans with extra limbs.
Latest comments
8 comments
Write a comment
Your email address will not be published. Required fields are checked *
Сергей
The new Stable Diffusion 3 Medium model from Stability AI turned out to be an interesting experiment, but with major drawbacks. Creepy three-armed and headless people are too much even for artificial intelligence! 😱
Анна
Yes, it’s strange to see such errors from the supposedly “most advanced” image generation model. Although, maybe this is just a troll in this way from Stability AI? 🤔
Франц
According to the information, the problem lies in limited training data, due to which the AI was unable to correctly “learn” human anatomy. Let's hope the developers will fix this in future versions of the model. 💪
Валерий
Something tells me that these “mutants” will become a new meme on the Internet! 😆 People always find a way to laugh at technical mistakes.
Отторино
What newfangled nonsense! 😒 Why do we even need these useless image generators? It would be better to do something useful for society.
Джулия
Ottorino, don't be such a grumbler! 🙃 Image generation technologies open up new opportunities for creativity, marketing and many other areas. It's worth giving them a chance to develop.
Франц
I completely agree with Julia. Despite some funny bugs, Stable Diffusion has a lot of potential! 💡 Developers are already experimenting with models for creating videos, animations and even program code.
Анна
By the way, did you hear that investors recently valued Stability AI at $1 billion? 💰 It looks like this industry is poised for serious growth in the coming years, despite some problems for the company itself.