Weekly Update - Meeting Hannah, CG and character commissions
Added 2023-03-10 15:34:32 +0000 UTCHey y'all!
I trust that most of my subscribers have met the new girl, Hannah. Here is an AI generated portrait of her made through the new service Holara AI. I've decided to subscribe to the service for a month to test it out. Here are my thoughts about it :
Holara seems determined to generate moe / little girl type characters, so it often feels like I am fighting against it. The prompts needed to communicate with the AI are much more rigid than on NovelAI, so I think I might stick with NAI for the time being. But the end product often looks pretty stunning on Holara, as this portrait of Hannah shows. That muscle definition and skin details are very convincing.
Also, she has six fingers on one hand, but, oh well. It's AI, what did you expect?
I have received an update from Ainemswork for the grapevine CG, and it is very close to being completed. It looks very, very good, along the lines of their more recent works for the game. I'm really lucky to be working with such a talented artist.
Next, I plan to commission them to do Hannah's final character sprite. Would you be interested in seeing the commission document I'm planning to send them? I was thinking of sharing it to subscribers, if you are curious to see more of the behind-the-scenes process. The document contains at least one spoiler about the character, however, so be warned.
---
Apart from that, I believe we are due for a poll soon. Should I let you decide what kind of CG you would like to see with Hannah? A feat of strength? A submission hold? An H scene? I'm always curious to see what you can come up with, as a community. Let's talk about it!
Until then, have a good weekend and stay safe out there!
Comments
Using Holara AI, the prompt was as follows : 1girl, solo, (muscular female)++, (female bodybuilder)++++, (pink playboy bunny suit with frills on the hips)++, long blonde hair, halfbody, standing, contrapposto++, wide surprised blue eyes, looking at viewer, grinning, thick muscular thighs, detailed muscles, (thick muscular arms)++, (plain white background)++, (white thighhighs with pink frills)+, (huge breasts with skindentation, pushed-up breasts, overflowing breasts)+, six pack abs, slim waist, hip bones, collarbone, (visual novel character sprite)+++ With Undesired Content set to : lowres, bad anatomy, bad hands, text, error, missingfingers, extra digit, fewer digits, cropped, worst quality, low quality,normal quality, jpeg artifacts, signature, watermark, username, (deformed), [blurry], bad anatomy, disfigured, poorly drawn face,mutation, mutated, (extra_limb), (ugly), (poorly drawn hands), messydrawing, broken legs, ribbons, gloves, wings, front bangs, messy hair, Then, it takes a LOT of generations to get one potent image. For example, the one attached to this post is great when you look at it quick, but it goes overboard with the frills, the fingers, and it kept wanted to make Hannah hold a bar or a stick or weights, it was tiring. Adding stuff in the Undesired Content field sometimes seem to make Holara prioritize adding it to the generated picture, so it feels quite broken. Overall I don't enjoy Holara quite as much as NovelAI for image generation. The prompts are not really well understood by its engine and it doesn't feel as natural as NovelAI overall. I do not know if I will renew my current subscription. But recently, Holara added an Image-to-Image function, and it makes the whole thing much better. Now my workflow looks more like : generate image on NovelAI -> Upload Novel AI image to Holara -> Generate in Holara using image-to-image strength of about 30-40. This workflow allows me to have the base generation of NovelAI and the upscaling and level of clear detail of Holara's Aika Aesthetic Anime model. I am also keeping an eye on Niji Journey, but last I checked it's still only public on their Discord, and I don't feel like practicing my prompting in public. I'd rather they release a private paid service I could try on my own time.
Peach Punch Games
2023-03-30 02:01:16 +0000 UTCVery good for IA rendering What prompt did you use ?
Frantz Lasvignes
2023-03-29 22:05:27 +0000 UTCi love this image
Marek Jarecki
2023-03-12 16:40:18 +0000 UTC
