V1.5
Read before:
this lora is trained only with fanarts so it will not have the resemblance to its anime or live action version, but it will keep the marks on the face, hair etc....
improved training
improved the flexibility of the lora
improved the marks on the face are now easier to generate
improved the clothes, now you can put on any haircut more easily
now you can do anime, 2.5D and even realistic style more easily
you can change leotard, for bodysuit to obtain a more similar clothing to the one in your movie.
you can use white humanoid robot and remove the clothing tags, to obtain a body similar to the one in the movie.
you can get different results depending on the sampler you use, my favorites are these:
that's all, enjoy!
V1.0
New LoRa this time from Gally (Alita in united states and many other countries)
This lora was trained with fanarts, so if you want a resemblance to the 2019 movie you should use the Alita lora, or if you prefer the anime one use alita ova (they are not mine),
The recommended weight is 0.7 to 1 and the recommended sampler is DDIM, this gives excellent results, the generated images were created with the following models: hassaku (hentai model), animatrix and anything v5 (with this was trained).
without more to write I hope you enjoy it and share your generations in the comments!