Both for convenience and performance, I like to combine my everyday LoRAs with my favourite models. HyperFlux Schnell is a blend of Flux Schnell with the HyperFlux LoRA at strength 0.12 and AntiBlur at strength 2.0.

Please respect the licenses of each of the creators whose dedicated work created these resources. For example, do not use HyperFlux Schnell on any service that monetizes image creation, such as for online image generation. Also, do not sell or license this model for a fee or something else of value.

Usage

My preferences for HyperFlux Schnell are a CFG of 3.5, but going up to 6.5 for special effects. I set sampling steps to 4, the Schnell default, and use the euler sampler with the simple scheduler.

The Q8 and Q6 variants were tested with the "t5xxl_fp16.safetensors" Clip Model, and the smaller quantizations were setup with the "t5xxl_fp8_e4m3fn.safetensors" Clip Model.

By default, HyperFlux Schnell creates images with a more or less deep depth of field. Put "bokeh" in the positive prompt to create a shallow depth of field.

If this model creates NSFW images it is not intentional.

Credits

Flux Schnell: https://civitai.com/models/618692?modelVersionId=699279
AntiBlur LoRA: https://civitai.com/models/675581/anti-blur-flux-lora
HyperFlux 8 LoRA: https://huggingface.co/ByteDance/Hyper-SD
This project was created using this wonderful resource:
https://civitai.com/articles/8322/merge-a-lora-into-flux-for-better-speed-and-quantize-it

Testing

Most of the time for this type of project goes into testing, both to establish an optimal balance between the LoRAs and to verify the finished model at the various quantizations. The default speed of Schnell image production was certainly appreciated. The complete test results for Q8 and Q5 at the Pure Fooocus Facebook group in the Base Model Guide.

Compared to the Q8, the Q5 produced less logical consistency in landscapes and bodily integrity in group subjects like a poker game. Portrait quality is essentially the same, but unless storage space is at a premium, I think the Q8 is a much better choice. And VRAM usage was only 0.3GB less for the Q5, which is what I would expect for a GGUF model.