ChromeShineXL

CHECKPOINT
Original


Updated:

67

Flavor-of-the-Month style mix. It's somewhat responsive to Pony-based loras but as it's a style mix, it's far enough diverged that most style loras will not have as good as an effect they would on Pony-V6 or more adjacent models like AutismMix which essentially just PonyXL but with smooth-anime lora merged in. Given that SD3 is soon to release next week, I doubt this will get much use anyway.

Unlike MIX-GEM-WXL, the unconditional prompting is optional to some degree, but you should probably try to uncond for tags like 'polydactyly' or 'extra digits'. I see a lot of people prompting for things like 'beautiful woman' or 'sensual expression' and I want to go ahead and say that that sort of prompting actually doesn't work at all. Prompting is entirely based on the tagging that goes into the image data and unless the trainer is tagging 'beautiful' or 'sensual' when they're preparing their training data, the AI will not learn the image associations of those words. If you want to understand how the CLIP TE that Stability uses understands images, look at this.

Obviously for Pony and other illustrative model, there is additional finetuning that's done. But the tagging there also based on set standards. Most training data is pulled from either Danbooru or e621. If you want to invoke training that has been done on training data pulled from there, you need to use their tags.

In any case do not use subjective terms when it comes to AI. Aside from quality grading tags like score_9, score_8, score_7 or best quality, very aesthetic that trainers use for bucketing purposes, the AI will not really understand what you're talking about because AI only knows what you've trained into it.

Anyway for this model, score tags other than score_9, score_8, score_7 don't really do much. You can throw them in, but they're kind of just a waste of tokens. Likewise I recommend using e621_p_low instead of score tags in negative.

ChromeLightXL is the Style Extract for this model. It can retain maybe around 55% to 70% of the model's style, from what I can tell, however it is much smaller and can be combined with various base models to take advantage of perhaps improved compositional sense. Up to you if you want to use it.

Version Detail

Pony
Flavor-of-the-Month style mix. It's somewhat responsive to Pony-based loras but as it's a style mix, it's far enough diverged that most style loras will not have as good as an effect they would on Pony-V6 or more adjacent models like AutismMix which essentially just PonyXL but with smooth-anime lora merged in. Uploaded onto tensorart so you can run it with their inference. For local, you can download it from the CivitAI page here: https://civitai.com/models/498178

Project Permissions

Reprinting is strictly prohibited

    Use Permissions

  • Use in TENSOR GREEN Online

  • As a online training base model on TENSOR GREEN

  • Use without crediting me

  • Share merges of this model

  • Use different permissions on merges

    Commercial Use

  • Sell generated contents

  • Use on generation services

  • Sell this model or merges

Comments

Related Posts

No posts yet
Describe the image you want to generate, then press Enter to send.