File size: 1,920 Bytes
d2055a0 bac7779 d2055a0 71dc4f5 8e34d37 d2055a0 4146ac7 ace393a 4146ac7 ace393a 04c2c12 f4bfd64 04c2c12 4146ac7 b2a33a3 4146ac7 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 |
---
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.
|