I didn’t expect it to take this long for me to get the new text model under control. It’s much more intelligent in general, but the repetition and random fixations can get surprisingly bad.
Sorry about that. This is just a heads up that I’m still working on it, and it’ll likely remain a bumpy ride for the next couple of months.
Thanks to those who have complained here and on reddit, especially the posts that give concrete examples of issues, and attempted workarounds, etc.
Please do keep posting any issues/discoveries, ideally with share links to give me specific examples that I can test.
Other news:
-
I’m still working on the next image gen update, and am hoping it will improve fine details, text rendering, and overall image quality. It should also be about 2x faster. If all goes well (which, from experience, it won’t) I should be able to get that done in the next month or so.
-
For a long time I’ve been very close to shipping a multiplayer plugin and database plugin. I’ve actually made multiple ‘ready to ship’ draft plugins of both, but have been disatisfied with the design and ended up not publishing each time. Using these plugins won’t trigger ads on the generator, so I need to be really careful about the overall architecture to ensure the the servers for them are dirt-cheap to run. I also want to make them fun/simple to use of course. Once I get text and images stable, I’m going to focus on these two plugins and finally get them done.
I lost my previous perchance account to lost login/pw credentials. I created it to follow perchance news. The last post by perchance was a few months ago… Are there any new revelations or movements in the text gen model? I think the “upgrade” was mentioned well over a year ago. Is it just too big a project for a one-person dev operation to handle?
Hello, with all due respect, AI Story Generator is totally terrible right now.
Right now the model is straight up nonsense, ingoring the whole plotline, outputting texts that has just no correlation to the previous ones, jumping scenes without being told to, or refusing to change a scene and follow the new one, etc. It ignores already existing context and details, also makes characters know things they very clearly should not. A few months ago it worked relatively fine, even with the new AI being sometwhat bad at storytelling, now it’s plainly dumb, ignoring the user, and even totally senseless.
We know, I hope this text model is fixed, I cannot stand the new model, this was a bad idea from the beginning
deleted by creator
I pray that this is only temporary, because for weeks now the image quality has been extremely poor, making the results unusable. Despite careful prompts and using the “Professional Photo” mode, human bodies are only rendered as 1950s pin-ups with uniform brown-beige plastic skin, and the faces all look the same, like sticker doll masks. Damn, it used to be so good and it was so interesting to create individual, distinctive human characters! I really hope that at least the previous quality will be restored! Thank you for your work!
Perchance images now look very different and low brightness. Why is Perchance update quality always in negative ??
I facing same issue in the last 48 hrs
so like the models censored now? bad look
So much ‘technical difficulties’ for a week now. Everything is so slow now. Please fix!
The new update only pays attention to the overview for the first two or three responses and then it’s like the overview doesn’t exist. It also doesn’t understand sensory detail. It has no concept of furniture, a room, the environment. It doesn’t follow writing styles. It doesn’t follow the details of the previous message. It went as far today as one paragraph in third person, the following paragraph in first person. It still defaults to clinical detatched language despite my robust instructions to embody a characters deep pov. It still writes like it’s following a checklist, lazy summaries to get from point A to point B of any paragraph. It still fights NSFW and the more you fight it the more it defaults to ‘safe’.
Why perchance developer is lying?? Fine detail and overall image quality didn’t improve but it only became worse. And it’s clearly not 2x faster only slightly faster. Text rendering may be better that is all. So now ai chat is very broken and ai image is very worse. Perchance’s update quality is in negative. 👎
to EVERYONE: perchance isn’t lying.now image quality fixed very well. but something habit of prompt changed. if you tired of CARTOONS, try copy these to your image prompt and put at the front :
“photorealistic, realistic,realistic person,realistic style.real photo,realistic photo,real person, photorealistic，hyperrealistic，real human portrait， candid photography，ultra-detailed skin texture，documentary style，natural body proportions， natural lighting.
then NO MORE CARTOONS. you are welcome.
image quality fixed how?? are you wearing blindfold??
already told you how,are you blind.haha
Tnx for the prompt. I’ve tried a few dozen renders and your ‘prefix’ does help a bit. I’ll try many more times with some permutations and such and see how it works. But still, it’s tricky and not really the same with some cartoonish elements showing through, like too much contrast on skin tones accentuating the muscles and such. And I used perchance for more academic artistic painting styles, not really for photos. Maybe I’ll get those back with some experimenting, but I bet not entirely if the underlying Stable Diffusion is no longer used.
But in any case, tnx for the tip. I appreciate it.
Repetitive, it’s not just x; it’s y, excessive semicolons, repetitive, Arthur is everywhere, repetitive, heavy and difficult to read and so fucking repetitive
Text model is horrible.
It will not stop defaulting to clinical terms like a medical journal and writing like a detatched romance novel. NSFW has just disappeared, become completely sanitized. I have changed the tips in the generator to REQUIRE NSFW, I have given strict instructions of how, I have used the tags ‘Uncensored’ in the overview and it generated this ‘their mouths finally met, and then there was no more thinking — just heat, slick skin, and the sound of breathing gone ragged in the dark.’ skipped to morning after. This model is broken.
still waiting on the text gens to finish syncing up of my old stories and the rpgs. image mdels are a little out of sync and try to MERGE characters with background, I prompted that NOT to happen but it did anyways and the video…are they still on the back burner?
any updates guys…
Lately yhe images generated are very poor . All my seed are not able to keep the same face or scenario
I’m not a fan of the current text gen. I want the old one back. It was way more fun to make NSFW stories with the old model.
After a week of improvement, since the downtime it has returned to rushing, evasive, flowery, clinical.
The model has been performing really poorly for the past month and a half.
Dear Dev, if you are reading this, I cannot continue my stories until I see a major improvement on the text model that happend to me and majority of us users back in September of last year, I have been using Perchance for nearly two years now, I am not being mean, I am just trying to be nice, I cannot stand the aggressiveness of the characters. Thank you.
Hard agree, text model for AI Story Generator is totally terrible.
I have stories from years, and right now it’s just straight up nonsense, even ingoring the whole plotline or outputting texts that has just no correlation to the previous ones. It jumps scenes without being told to, or totally refuses to change a scene and follow the new one. It ignores context and details, and makes characters know things they very clearly should not. A few months ago it worked relatively fine, even with the new AI being sometwhat bad at storytelling, now it’s plain dumb and even senseless.