FLUX.1 Dev LoRa aiming to generate Wimmelbilder Where's Waldo style images

Tips

• Describe your scene naturally

• Trigger might be necessary but always test without

• Strength at 1.0 without trigger and 0.7/0.8 with trigger

• This model really likes landscape aspect ratios

• At 1.0 strength, the style is way better, but the model tends to generate a lot of malformed people

• If you really want to get the best quality possible, try generating at 1664x1216

How to use it

• Used best with a Comfy fp16/bf16 checkpoint, use it as any other Lora

• Use it with FAL FLUX with Loras: https://fal.ai/models/fal-ai/flux-general . Just point to the .safetensors file in this repo

How to get the best results

• Generate your base image with txt2img

• Upscale with something like AuraSR

• Do a second img2img pass with ~0.3-0.5 denoising strength, same prompt

• At the time of this release, tiled upscale is not possible yet with FLUX, but that would be really beneficial for this kind of outputs

Contact

https://twitter.com/nerijs