Clone Wars Style for Anima
This is my first Clone Wars Style checkpoint trained for the Anima model.
Unlike my previous release, which was trained on Illustrious, this version has been built entirely for the Anima ecosystem to better reproduce the stylized CGI look of Star Wars: The Clone Wars.
This project is still a work in progress, and there is still a lot I want to improve. While some characters, species, and environments already produce results that I'm very happy with, others still require additional training and a larger dataset to fully capture the unique visual identity of the series.
Even so, I'm genuinely proud of what this checkpoint has already achieved. It successfully reproduces much of the distinctive Clone Wars atmosphere, including its stylized CGI materials, lighting, color palette, character proportions, and cinematic presentation.
My long-term goal is to create a complete Clone Wars checkpoint capable of generating characters, environments, creatures, vehicles, and original scenes from across the Star Wars universe without relying on character-specific LoRAs.
This version includes
First Clone Wars checkpoint trained specifically for Anima
Improved stylized CGI rendering
Better Clone Wars-inspired lighting and materials
More consistent character proportions
Improved facial rendering
Expanded dataset featuring heroes, villains, aliens, creatures, droids, vehicles, and environments
Better cinematic compositions
Some characters are naturally more accurate than others. Complex non-humanoid characters and certain species will continue to improve as the dataset grows and future versions are released.
This checkpoint is intended for creating both faithful recreations of scenes from the animated series and entirely original Star Wars artwork while preserving the iconic Clone Wars visual style.
Thank you to everyone who tests my models and shares feedback. Every release helps this project move one step closer to my ultimate goal.
May the Force be with you.
Description
Comments (2)
What tooling did you use to train this? Lot's of people have shared how they train LORA but a checkpoint finetune like this is very interesting and I would love it if you shared a little about the process. Thanks!
I trained this checkpoint using Anima Standalone Trainer .
The training was done as a full checkpoint fine-tune on the official Anima base model. My dataset is entirely hand-curated and captioned, with a strong focus on character consistency and style. I spent a lot of time refining the captions and removing poor-quality images to keep the dataset as clean as possible.
I'm still experimenting with the training settings.
I also found the software to be very stable. Compared to kohya_ss, where I often experienced large fluctuations in training speed, Anima Standalone Trainer maintained a much more consistent performance. It's also very easy to use while still giving good control over the different training options and parameters.