I thought that OpenAI had this long-standing challenge for AI image models mastered…
Fingers. 3 of 3 are goofy.
High water mark of alternate AI (that will make happy non-plastic children).
I thought that OpenAI had this long-standing challenge for AI image models mastered…
Fingers. 3 of 3 are goofy.
High water mark of alternate AI (that will make happy non-plastic children).
Yes, I discovered it today too. In addition to the cue words, these have also started to appear. I am starting to not understand the cue words.
The image was generated at PR16. This is normal for this clunky model.
The problem with fingers, snakes, tentacles, etc. is a whole different issue, it doesn’t directly relate to the quality reduction of the latest update, unless it has perhaps gotten a bit worse. The current stable diffusion models seem to have issues with topology. Snakes in the shape of a ring or pieces floating freely in the air… there is still a lot that can be improved.
In PR16, the problem is not only with quality, but also with the level of knowledge and perception. It is better to be patient for 3 weeks and not to spoil your nerves. At least in PR13 the problems with hand generation are minimal. In that version only sometimes incorrectly or not there draws a hand or foot. But the chance of such an error is 1/500.
If it returns to the old model, which would make all of us happy, the greatest credit belongs to @kaienkala and @wincupgo . Your contribution is significant, thank you.
Bing Image Creator doesn’t understand prompt words right now. This is annoying. But maybe it’s because they are adjusting back to PR13.
This is everyone’s work, not mine. I just put together a timeline of the failure and hypothesized what the problem was.
Ribas answered almost all questions on DALL-E. And there were a lot of complaints.
Asked why GPT uses PR16 and not PR13
Hi there, The specific versioning and naming conventions, such as “PR13” and “PR16,” related to DALL·E 3 or any other models, are internal details that typically pertain to updates, improvements, or specific configurations of the models. OpenAI continuously works on improving the capabilities, efficiency, and accuracy of its models, including DALL·E 3. Updates or iterations like moving from “PR13” to “PR16” could involve enhancements in image generation quality, speed, or other performance metrics. It’s important to note that OpenAI aims to provide users with the best possible experience and results. Therefore, if GPT is working with a version labeled as “PR16” instead of “PR13,” it likely means that “PR16” includes updates or improvements that enhance the model’s performance or capabilities compared to “PR13.” For the most accurate and detailed explanation regarding the specific versions and their differences, it would be best to refer directly to OpenAI’s official announcements or documentation, as they provide insights into updates and improvements made to their models. If you have any more questions or need further assistance, feel free to ask. Best,
OpenAI Team
It seems that the situation is getting better, but it will take some time.
It looks like the malfunction started to get worse today. It’s like it’s never been fixed.
The corrections started most likely on 29 December. I had no problems with generations then. I checked a lot of queries and there were good results everywhere.
But these changes did not affect all users. And then they probably limited these changes to subscription holders, and gradually started to expand the user base.
Considering how it switches to PR16 when the boost is depleted, is it correct to understand that PR13 is a concept quality model for demonstration purposes and PR16 is an economy performance model?