Sammy Krea v5 LoRAs
Selected production candidates from the Sammy Krea v5 checkpoint evaluation:
- Character identity: step 800, trigger
sammy
- General visual style: step 1400, trigger
sammy style
Fal-compatible files
Use the files in fal-compatible/ with fal-ai/krea-2/turbo/lora. They use
Diffusers-style Krea 2 LoRA keys and were verified through Fal's queue API.
fal-compatible/sammy-character-krea2-v5-step800.safetensors
- SHA256:
2427e1344033434a0e18e81f11c01762c947b9464480c71db4b2583b18a810e6
fal-compatible/sammy-style-krea2-v5-step1400.safetensors
- SHA256:
551d60ec904e9c677750dcad97143cb563c712bf33fb2967f7d846126176dc2b
The root-level checkpoints are the original musubi-tuner outputs. They work in
the native musubi inference pipeline but must not be passed directly to Fal's
managed Krea 2 endpoint. Fal can accept those URLs without applying the weights.
The adapter weights are subject to the Krea 2 base model's community license and
any applicable rights in the training material.
