README.md

---
license: other
license_name: flux-1-dev-non-commercial-license
license_link: https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/LICENSE.md
base_model:
- black-forest-labs/FLUX.1-dev
---

<img src="https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/resolve/main/images/banners/ComfyUI_2024-11-30_0088.jpg" alt="FLUX">

**===========================================**

**FLUX Model Merges & Tweaks: Detail Enhancement and Acceleration**

<img src="https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/resolve/main/images/banners/ComfyUI_2024-11-30_0093.jpg" alt="FLUX Model Merges & Tweaks: Detail Enhancement and Acceleration">

This repository contains merged models, built upon the base models:
- [Freepik's Flux.1-Lite-8B-alpha](https://huggingface.co/Freepik/flux.1-lite-8B-alpha)
- [Flux-dev-de-distill](https://huggingface.co/nyanko7/flux-dev-de-distill)

Detail enhancement and acceleration techniques have been applied, particularly optimized for NVIDIA 4XXX cards (maybe 3XXX too). The goal is to have high efficiency accelerated models with lower overhead. The (de-re-destill) model can be used with CFG remaining at 1. Also the baked-in accelerators work as intended.

**===========================================**

<img src="https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/resolve/main/images/banners/ComfyUI_2024-11-30_0097.jpg" alt="Detail Plus!">

**Detail Plus! - Built upon the base model [Freepik's Flux.1-Lite-8B-alpha](https://huggingface.co/Freepik/flux.1-lite-8B-alpha):**

**Detail Enhancement Used:**
- **Style LORA - Extreme Detailer for FLUX.1-dev** (Weight: 0.5) ([Model Link](https://civitai.com/models/832683))
- **Best of Flux: Style Enhancing LoRA** (Weight: 0.25) ([Model Link](https://civitai.com/models/821668))

1. **SAFETensors Format (fp8_34m3fn_fast)**:
   - [flux.1-lite-8B-alpha-Hyper-8.Steps-Detail.Plus-fp8_e4m3fn_fast](https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/flux.1-lite-8B-alpha-Hyper-8.Steps-Detail.Plus-fp8_e4m3fn_fast.safetensors)
   - [flux.1-lite-8B-alpha-Hyper-16.Steps-Detail.Plus-fp8_e4m3fn_fast](https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/flux.1-lite-8B-alpha-Hyper-16.Steps-Detail.Plus-fp8_e4m3fn_fast.safetensors)
   - [flux.1-lite-8B-alpha-Turbo-8.Steps-Detail.Plus-fp8_e4m3fn_fast](https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/flux.1-lite-8B-alpha-Turbo-8.Steps-Detail.Plus-fp8_e4m3fn_fast.safetensors)

2. **SAFETensors Format (FULL)**:
   - flux.1-lite-8B-alpha-Hyper-8.Steps-Detail.Plus.safetensors
   - flux.1-lite-8B-alpha-Hyper-16.Steps-Detail.Plus.safetensors
   - flux.1-lite-8B-alpha-Turbo-8.Steps-Detail.Plus.safetensors

3. **GGUF Quantized Models (Q8_0)**:
   - [flux.1-lite-8B-alpha-Hyper-8.Steps-Detail.Plus.Steps.Q8_0_quantized](https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/flux.1-lite-8B-alpha-Hyper-8.Steps-Detail.Plus.Steps.Q8_0_quantized.gguf)
   - [flux.1-lite-8B-alpha-Hyper-16.Steps-Detail.Plus.Steps.Q8_0_quantized](https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/flux.1-lite-8B-alpha-Hyper-16.Steps-Detail.Plus.Steps.Q8_0_quantized.gguf)
   - [flux.1-lite-8B-alpha-Turbo-8.Steps-Detail.Plus.Steps.Q8_0_quantized](https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/flux.1-lite-8B-alpha-Turbo-8.Steps-Detail.Plus.Steps.Q8_0_quantized.gguf)

**===========================================**

<img src="https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/resolve/main/images/banners/ComfyUI_2024-11-30_0119.jpg" alt="Detail Plus! De-Re-Distilled">

**Detail Plus! De-Re-Distilled (Built upon the base model [Flux-dev-de-distill](https://huggingface.co/nyanko7/flux-dev-de-distill)

**Detail Enhancement Used:**
- **Style LORA - Extreme Detailer for FLUX.1-dev** (Weight: 0.15) ([Model Link](https://civitai.com/models/832683))
- **Best of Flux: Style Enhancing LoRA** (Weight: 0.15) ([Model Link](https://civitai.com/models/821668))

**Re-Distillation Used:**
- **Flux distilled lora** (Weight: -1.00) ([Model Link](https://civitai.com/models/977247/flux-distilled-lora))

1. **SAFETensors Format (fp8_34m3fn_fast)**:
   - [Flux-dev-de-re-distill-Hyper.8.Step-Detail.Plus-ReDis-fp8_e4m3fn_fast](https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/de-re-distill/FP8/Version%202/Flux-dev-de-re-distill-Hyper.8.Step-Detail.Plus-ReDis-fp8_e4m3fn_fast-V2.safetensors)
   - [Flux-dev-de-re-distill-Hyper.16.Step-Detail.Plus-ReDis-fp8_e4m3fn_fast](https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/de-re-distill/FP8/Version%202/Flux-dev-de-re-distill-Hyper.16.Step-Detail.Plus-ReDis-fp8_e4m3fn_fast-V2.safetensors)
   - [Flux-dev-de-re-distill-Turbo.8.Step-Detail.Plus-ReDis-fp8_e4m3fn_fast](https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/de-re-distill/FP8/Version%202/Flux-dev-de-re-distill-Turbo.8.Step-Detail.Plus-ReDis-fp8_e4m3fn_fast-V2.safetensors)

2. **SAFETensors Format (FULL)**:
   - Flux-dev-de-re-distill-Hyper.8.Step-Detail.Plus-ReDis-MAIN-V2.safetensors
   - Flux-dev-de-re-distill-Hyper.16.Step-Detail.Plus-ReDis-MAIN-V2.safetensors
   - Flux-dev-de-re-distill-Turbo.8.Step-Detail.Plus-ReDis-MAIN-V2.safetensors

3. **GGUF Quantized Models (BF16)**:
   - Flux-dev-de-re-distill-Hyper.8.Step-Detail.Plus-ReDis-MAIN-V2-BF16.gguf
   - Flux-dev-de-re-distill-Hyper.16.Step-Detail.Plus-ReDis-MAIN-V2-BF16.gguf
   - Flux-dev-de-re-distill-Turbo.8.Step-Detail.Plus-ReDis-MAIN-V2-BF16.gguf

4. **GGUF Quantized Models (Q8_0)**:
   - [Flux-dev-de-re-distill-Hyper.8.Step-Detail.Plus-ReDis-MAIN-V2-Q8_0](https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/de-re-distill/GGUF/Q8_0/Flux-dev-de-re-distill-Hyper.8.Step-Detail.Plus-ReDis-MAIN-V2-Q8_0.gguf)
   - [Flux-dev-de-re-distill-Hyper.16.Step-Detail.Plus-ReDis-MAIN-V2-Q8_0](https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/de-re-distill/GGUF/Q8_0/Flux-dev-de-re-distill-Hyper.16.Step-Detail.Plus-ReDis-MAIN-V2-Q8_0.gguf)
   - [Flux-dev-de-re-distill-Turbo.8.Step-Detail.Plus-ReDis-MAIN-V2-Q8_0](https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/de-re-distill/GGUF/Q8_0/Flux-dev-de-re-distill-Turbo.8.Step-Detail.Plus-ReDis-MAIN-V2-Q8_0.gguf)

5. **GGUF Quantized Models (Q6_K)**:
   - Flux-dev-de-re-distill-Hyper.8.Step-Detail.Plus-ReDis-MAIN-V2-Q6_K.gguf
   - Flux-dev-de-re-distill-Hyper.16.Step-Detail.Plus-ReDis-MAIN-V2-Q6_K.gguf
   - Flux-dev-de-re-distill-Turbo.8.Step-Detail.Plus-ReDis-MAIN-V2-Q6_K.gguf

6. **GGUF Quantized Models (Q4_K_S)**:
   - Flux-dev-de-re-distill-Hyper.8.Step-Detail.Plus-ReDis-MAIN-V2-Q4_K_S.gguf
   - Flux-dev-de-re-distill-Hyper.16.Step-Detail.Plus-ReDis-MAIN-V2-Q4_K_S.gguf
   - Flux-dev-de-re-distill-Turbo.8.Step-Detail.Plus-ReDis-MAIN-V2-Q4_K_S.gguf

**===========================================**

**Acceleration Credits:**

* [ByteDance](https://huggingface.co/ByteDance), a leading AI technology company, optimized the following models:
   - [Hyper-FLUX.1-dev-16steps](https://huggingface.co/ByteDance/Hyper-SD/blob/main/Hyper-FLUX.1-dev-16steps-lora.safetensors)
   - [Hyper-FLUX.1-dev-8steps](https://huggingface.co/ByteDance/Hyper-SD/blob/main/Hyper-FLUX.1-dev-8steps-lora.safetensors)

* [Alimama Creative](https://huggingface.co/alimama-creative), a renowned NLP innovator, optimized the following model:
   - [FLUX.1-Turbo-Alpha](https://huggingface.co/alimama-creative/FLUX.1-Turbo-Alpha)

**===========================================**

**Attribution and Licensing Notice:**

The [FLUX.1-dev Model](https://huggingface.co/black-forest-labs/FLUX.1-dev) is licensed by Black Forest Labs, Inc. under the FLUX.1-dev [Non-Commercial License](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md). Copyright Black Forest Labs, Inc.

Our model weights are released under the FLUX.1-dev [Non-Commercial License](https://huggingface.co/mdmachine/FLUX.Model.Merge-Detail.Enhancement.and.Acceleration/blob/main/LICENSE.md).

This merge combines the strengths of these models, applying detail enhancement and acceleration techniques to create a unique and powerful AI model built upon [Freepik's Flux.1-Lite-8B-alpha](https://huggingface.co/Freepik/flux.1-lite-8B-alpha) & [Flux-dev-de-distill](https://huggingface.co/nyanko7/flux-dev-de-distill). We hope this contributes positively to the development of the NLP community!