LoRA for the 10Eros‑Max
This LoRA was created by simply extracting the weight difference between 10Eros_Max_h3_TURBO-hybrid_beta5 and 10Eros_Max_h3_hybrid_beta5.
The goal is simple: to enable 10Eros_Max_h3_hybrid_beta5 to work with fewer steps (14–16) while maintaining audio quality.
How to use
Just apply this LoRA (rank 128) with a weight of 0.4–0.6 to the 10Eros_Max_h3_hybrid_beta5 model and use "spectrum" to achieve a good balance between image/audio quality and generation speed."Res_multistep + Simple" is a good starting point.
Credits
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support