Not long ago, StabilityAI released the open source StableDiffusion 3Medium, which made netizens exploring AI very excited. After all, the release of SD1.5 and SDXL has brought a lot of shock. This time, SD3 is even more exciting. Many netizens are excited. I downloaded it as soon

entertainment 9970℃
Not long ago,

, stabilityai released the open source stablediffusion 3medium, which made netizens exploring AI very excited. After all, the release of sd1.5 and sdxl has brought a lot of shock. This time, sd3 is even more anticipated. Many netizens They all downloaded it as soon as it was open sourced, but the experience did not seem to be as great as stabilityai said.

Not long ago, StabilityAI released the open source StableDiffusion 3Medium, which made netizens exploring AI very excited. After all, the release of SD1.5 and SDXL has brought a lot of shock. This time, SD3 is even more exciting. Many netizens are excited. I downloaded it as soon - Lujuba

In the initial publicity, stabilityai said that sd3 can already generate text, but currently it can only generate English, and Chinese cannot be recognized, and most importantly, sd3 seems to have gone backwards in character generation. For the generation of human portraits, SD3 often produces deformed character forms, especially errors in human limbs, hands and feet, and the overall experience is not as good as the previous generation model.

Not long ago, StabilityAI released the open source StableDiffusion 3Medium, which made netizens exploring AI very excited. After all, the release of SD1.5 and SDXL has brought a lot of shock. This time, SD3 is even more exciting. Many netizens are excited. I downloaded it as soon - Lujuba

In other words, as long as the limbs of the characters are not generated, the performance of the sd3 model on the local screen is still very exquisite. In the generation of objects and animals, sd3 basically has no problems, and it also has a high performance for complex long prompt words. The recognition rate is good, but it cannot draw human beings with sound limbs. In this regard, it has degraded to average, which has caused confusion among many people.

Not long ago, StabilityAI released the open source StableDiffusion 3Medium, which made netizens exploring AI very excited. After all, the release of SD1.5 and SDXL has brought a lot of shock. This time, SD3 is even more exciting. Many netizens are excited. I downloaded it as soon - Lujuba

foreign forum reddit users analyzed that the error in understanding the human body limbs in SD3 was because during training, developers interfered too much with the content, set filtering restrictions for content, and the conditions for passing it were very harsh, which led to AI recognition errors. Some Normal pictures are therefore recognized as restricted content. Therefore, in the public sd3 model, the effect of the human body is deformed due to insufficient training.

Not long ago, StabilityAI released the open source StableDiffusion 3Medium, which made netizens exploring AI very excited. After all, the release of SD1.5 and SDXL has brought a lot of shock. This time, SD3 is even more exciting. Many netizens are excited. I downloaded it as soon - Lujuba

This explains why sd3 has not caused a big stir, because in addition to recognizing English, the most important human body recognition has gone backwards. It was originally solving the problem of hands, but now it has evolved into solving the problem of limbs. The problem requires a long prompt word to fix, but for non-human beings, the effect of sd3 is surprisingly good, the details are in place, and the rules of physics are followed. Unfortunately, this version is not perfect.

Not long ago, StabilityAI released the open source StableDiffusion 3Medium, which made netizens exploring AI very excited. After all, the release of SD1.5 and SDXL has brought a lot of shock. This time, SD3 is even more exciting. Many netizens are excited. I downloaded it as soon - Lujuba

Of course, this may be related to stabilityai's difficulties. Before that, the company's financial situation was not good, and the CEO resigned, and the former key members also left. Coupled with the layoffs, it is not easy to launch SD3 now. Some netizens said that the company There is internal chaos, as evidenced by the immature sd3, but in the end it was launched and open sourced. This has kept its previous promises. Maybe it is just a simple mistake? (The picture shows the stablediffusion 3 model architecture)

Not long ago, StabilityAI released the open source StableDiffusion 3Medium, which made netizens exploring AI very excited. After all, the release of SD1.5 and SDXL has brought a lot of shock. This time, SD3 is even more exciting. Many netizens are excited. I downloaded it as soon - Lujuba

But this is only a medium cup of sd3, which means that large cups and small cups will be launched in the future. Perhaps after experiencing this problem, stabilityai will modify the training data. It is possible to improve a better sd3, and there are also sd1.5 and sdxl that are being continuously created and used. If the modified sd3 can make further progress on the basis of sdxl, it is worthy of appreciation. Of course, if sd3 supports other languages It would be better to understand

Not long ago, StabilityAI released the open source StableDiffusion 3Medium, which made netizens exploring AI very excited. After all, the release of SD1.5 and SDXL has brought a lot of shock. This time, SD3 is even more exciting. Many netizens are excited. I downloaded it as soon - Lujuba

At present, the sd3 model only supports comfyui for the time being. Netizens who want to experience it can download comfyui and try it. Although the open source version of sd3 is a medium version, for the requirements of the graphics card, it is still recommended to have a larger video memory. OK, so we recommend using the GALAXY rtx4070 ti super Xingyao OC graphics card.

Not long ago, StabilityAI released the open source StableDiffusion 3Medium, which made netizens exploring AI very excited. After all, the release of SD1.5 and SDXL has brought a lot of shock. This time, SD3 is even more exciting. Many netizens are excited. I downloaded it as soon - Lujuba

GALAXY rtx4070 ti super Xingyao oc graphics card, compared to the 4070ti, the video memory is larger 16gbgddr6x, and the core performance has also been enhanced. Both AI creators and gamers can feel the obvious improvement. At the same time, 16gb The video memory is also very suitable for running AI software. It is a high-performance and large video memory graphics card that should not be missed. If you are interested in this graphics card, you are welcome to go to the GALAX official store to choose~

(8801617)

Tags: entertainment