4.1 is here, 22 more images for more resolution flexibility . Z-Image LoRa coming up!

V4 is here, and Tori has fully shifted to Qwen 🌟

Trained the model with 1 repeat per epoch, 44 images total of different resolutions. Turned out okay in my opinion, although occasionally messes up anatomy when given confusing prompts.

V 3.0 for Qwen release log:

Better at replicating the red leather outfit. Now with less artifacts and better hands. Flux remains the main model for Tori and most of my characters, however.

V 3.0 release log:

New dataset, remade with Chroma and Krea, upscaled and added ADetailer on her face, reduced artifacts on her hands. Although imperfectly, the model learned the red leather outfit from the Terminator 3 movie. To recreate it, refer to the clothes as "the canon red leather T-X outfit".

Happy creating! 🩷💋

V 2.0 release log:

Much more realistic and dynamic, with an original dataset, enhanced through Chroma.

V 1.0 release log:

Tori finally gone independent 👽❤️‍🔥💋

I tried to make her design as reminiscent of her inspirations as possible, while still introducing a lot of freshness and potential for use in very dynamic shots. And no, the model didn't learn the TX's original outfit. For that, please use my separate TX LoRa here:

https://tensor.art/models/831307812517820122?source_id=kT6_o1_ql0S3oPIubHb99REi