|
--- |
|
license: other |
|
language: |
|
- en |
|
base_model: |
|
- black-forest-labs/FLUX.1-dev |
|
- fal/AuraFlow-v0.3 |
|
pipeline_tag: text-to-image |
|
tags: |
|
- AuraFlow |
|
- Flux |
|
- FluxDev |
|
- Merge |
|
- AuraFlowPipeline |
|
- image-generation |
|
widget: |
|
- text: >- |
|
Test Prompt |
|
output: |
|
url: AuraFlux2test.png |
|
--- |
|
<Gallery /> |
|
|
|
A Diffusers conversion of an experimental model merging: |
|
[FLUX.1-Dev](https://huggingface.co/black-forest-labs/FLUX.1-Dev) by [Black Forest Labs](https://blackforestlabs.ai/announcing-black-forest-labs/). |
|
and |
|
[AuraFlow](https://huggingface.co/fal/AuraFlow-v0.3/) by [Fal.Ai](https://blog.fal.ai/auraflow/) and, namely, [@cloneofsimo](https://twitter.com/cloneofsimo) + [@isidentical](https://twitter.com/isidentical). |
|
|
|
The merging process was not conducted by me. |
|
I'm merely curating this conversion to make it usable/testable/trainable via Diffusers. |
|
A [repo](https://huggingface.co/NeuroDonu/flux-and-aura-merge) by [NeuroDonu](https://huggingface.co/NeuroDonu) hosts the very hefty Safetensors format merged model [file](https://huggingface.co/NeuroDonu/flux-and-aura-merge/blob/main/flux_and_aura_merge.safetensors). |
|
Presumably, NeuroDonu is also the original creator of this merge. |
|
If so, then please credit them if making further model modifications. |
|
|
|
I haven't yet had time to conduct any extensive enough comparative testing on this merge. |
|
But so far it appears to work more or less on point. |
|
And even seems to work with Flux LoRAs. |
|
|
|
You may try it out at [one of my custom LoRA galleries](https://huggingface.co/spaces/AlekseyCalvin/soonfactory) currently running this merge. |
|
|
|
 |
|
|
|
Notably, this merge seems to have retained at least some of AuraFlow's greater base-level comprehension of named artist styles. |
|
To be clear, I thus far have very little sense as to the degree, specificities, or limits of this. |
|
May this become more apparent with testing and use. |
|
|
|
 |
|
|
