OLOID Framework is an open-source LoRA framework for automotive design generation.
Built on Flux.1 dev, it is designed for agent retrieval and invocation, aiming to hand over the tedious parameter-tuning process in design gacha (LoRA stacking + seed sampling) workflows to agents and free up designers’ time.
It also supports manual combination and use by designers.
Framework Structure
The current release includes 49 LoRAs, organized across four dimensions and designed to be combined based on different design goals.
For example: PRD (define volume) + CRT (add form language) + INS (expand creative possibilities) + VIS (control visual output)
Dimension
Code
Role
Production
PRD
Production-oriented volume, proportion, and structural foundation
Creative
CRT
Advances form language and design variation
Inspiration
INS
Introduces cross-domain design language and additional stimulus
Visuals
VIS
Controls image style, material feel, and overall visual impression
A full structural overview is shown below:
Structure
Figure: Full structure across PRD / CRT / INS / VIS and EXT / INT / PART.
Index
For each LoRA’s trigger words, reference prompts, and recommended weights, please refer to the index:
The current phase of OLOID on Flux.1 has been completed. We have already moved on to exploring next-generation model training and a more complete agent-oriented workflow, so this generation will no longer be updated.
Feedback and testing are welcome.
For discussions around AI generation systems for automotive design, please contact OkeyDesign: hello@okeydesign.com
For custom training, agent-related and other technical topics, please contact XonarLabs: info@xonarlabs.com
License
This model is built on FLUX.1 dev.
Usage may be subject to the base model's license restrictions.
