Looking for AI Text to Image Generation Expert. I need Consistent Anime Characters Art
Budget: $30 – $250 USD
!!! You should be able to generate the exact same character in many poses. !!!
The goal is to create unique anime characters similar to Genshin Impact characters. Then create many poses for each character.
Each character should have many poses, some poses without and some with equipment/weapons present (such as swords, staff, guns, etc).
These characters will be used in new games, we will be making.
Examples (This isn't perfect similarity):
https://cdn.discordapp.com/attachments/1233056723246256273/1249429587968917597/1girl_short_hair_twintails_spiked_hair_blonde_hair_blue_eyes_skinny_small_s-355270157.png?ex=666d343b&is=666be2bb&hm=a5f9b0b7fdb6f5d5acc48c209b8487a8c356801faf4bf7932b173e19ed532fd5&
https://cdn.discordapp.com/attachments/1233056723246256273/1249429588325564496/1girl_short_hair_twintails_spiked_hair_blonde_hair_blue_eyes_skinny_small_s-574002293.png?ex=666d343b&is=666be2bb&hm=df05578a3e0b485c0d9f2d36c08c84dbdaa5f78a9c4e4e7f61688da45e8f29e3&
https://cdn.discordapp.com/attachments/1233056723246256273/1249429588711309452/1girl_short_hair_twintails_spiked_hair_blonde_hair_blue_eyes_skinny_small_s-1270271234.png?ex=666d343b&is=666be2bb&hm=70b7360b742132205793613a359b548b6aba9cae2075fc92267887aa648995db&
More Character Examples:
https://cdn.discordapp.com/attachments/1233056723246256273/1243258014249652315/1111.png?ex=666d2a01&is=666bd881&hm=0c20d387c9beb559ca5ee06370cb09866f904c0c6eca87e828f0f47b3f25bad4&
https://cdn.discordapp.com/attachments/1233056723246256273/1243258014622941214/2222.png?ex=666d2a01&is=666bd881&hm=dcd46fe45bb64040e157f9f099f41378751ed5433d568b5f11288f8efa0780f5&
https://cdn.discordapp.com/attachments/1233056723246256273/1243258014946033726/333333.png?ex=666d2a01&is=666bd881&hm=b11ecdffe179818364424a3b01eedb349c6f46dca9d0c714ddbbe92cb0a6e6a8&
------------
Some info about image achieving consistency I found:
1. Character Reference: https://www.christytuckerlearning.com/create-consistent-character-images-in-midjourney/ and https://petapixel.com/2024/03/12/you-can-now-recreate-the-same-character-across-different-ai-images-on-midjourney/
2. Fast-training the model on existing images: https://creator.nightcafe.studio/my-models
3. Set of views of the same character (place all ControlNet frames on one image.): https://onceuponanalgorithm.org/consistent-ai-character-generation-with-different-poses-in-stable-diffusion/
4. NovelAI Tagging+Vibe Transfer+Base Example Image: https://docs.novelai.net/image/tutorial-charactercreation.html -> This seems to be the most promising, doing EXACTLY what we need.
As far as I understand NovelAI trained their model on anime specifically and they also trained it to strictly follow pre-defined tags, so the character ends up being consistent across many images if you input the exact same tags into the prompts (https://docs.novelai.net/image/tutorial-charactercreation.html). It's not 100% consistent but it's really close. You might have to generate many images and select the ones that are very close.
------------
-Please write down whether you are able to generate consistent characters. (You could even show examples.)
-I will ignore stock bids.
-I will let the bidding run for a day.
The goal is to create unique anime characters similar to Genshin Impact characters. Then create many poses for each character.
Each character should have many poses, some poses without and some with equipment/weapons present (such as swords, staff, guns, etc).
These characters will be used in new games, we will be making.
Examples (This isn't perfect similarity):
https://cdn.discordapp.com/attachments/1233056723246256273/1249429587968917597/1girl_short_hair_twintails_spiked_hair_blonde_hair_blue_eyes_skinny_small_s-355270157.png?ex=666d343b&is=666be2bb&hm=a5f9b0b7fdb6f5d5acc48c209b8487a8c356801faf4bf7932b173e19ed532fd5&
https://cdn.discordapp.com/attachments/1233056723246256273/1249429588325564496/1girl_short_hair_twintails_spiked_hair_blonde_hair_blue_eyes_skinny_small_s-574002293.png?ex=666d343b&is=666be2bb&hm=df05578a3e0b485c0d9f2d36c08c84dbdaa5f78a9c4e4e7f61688da45e8f29e3&
https://cdn.discordapp.com/attachments/1233056723246256273/1249429588711309452/1girl_short_hair_twintails_spiked_hair_blonde_hair_blue_eyes_skinny_small_s-1270271234.png?ex=666d343b&is=666be2bb&hm=70b7360b742132205793613a359b548b6aba9cae2075fc92267887aa648995db&
More Character Examples:
https://cdn.discordapp.com/attachments/1233056723246256273/1243258014249652315/1111.png?ex=666d2a01&is=666bd881&hm=0c20d387c9beb559ca5ee06370cb09866f904c0c6eca87e828f0f47b3f25bad4&
https://cdn.discordapp.com/attachments/1233056723246256273/1243258014622941214/2222.png?ex=666d2a01&is=666bd881&hm=dcd46fe45bb64040e157f9f099f41378751ed5433d568b5f11288f8efa0780f5&
https://cdn.discordapp.com/attachments/1233056723246256273/1243258014946033726/333333.png?ex=666d2a01&is=666bd881&hm=b11ecdffe179818364424a3b01eedb349c6f46dca9d0c714ddbbe92cb0a6e6a8&
------------
Some info about image achieving consistency I found:
1. Character Reference: https://www.christytuckerlearning.com/create-consistent-character-images-in-midjourney/ and https://petapixel.com/2024/03/12/you-can-now-recreate-the-same-character-across-different-ai-images-on-midjourney/
2. Fast-training the model on existing images: https://creator.nightcafe.studio/my-models
3. Set of views of the same character (place all ControlNet frames on one image.): https://onceuponanalgorithm.org/consistent-ai-character-generation-with-different-poses-in-stable-diffusion/
4. NovelAI Tagging+Vibe Transfer+Base Example Image: https://docs.novelai.net/image/tutorial-charactercreation.html -> This seems to be the most promising, doing EXACTLY what we need.
As far as I understand NovelAI trained their model on anime specifically and they also trained it to strictly follow pre-defined tags, so the character ends up being consistent across many images if you input the exact same tags into the prompts (https://docs.novelai.net/image/tutorial-charactercreation.html). It's not 100% consistent but it's really close. You might have to generate many images and select the ones that are very close.
------------
-Please write down whether you are able to generate consistent characters. (You could even show examples.)
-I will ignore stock bids.
-I will let the bidding run for a day.