Flux2-Klein-9B-True-V3

如果您喜欢我的模型，请支持 ❤️ 我以获得更多的算力来调优模型。(请从 Huggingface 下载)

If you like my models, please support ❤️ me for more computing resources to do finetuning work. (Please download it from Huggingface)

=============================================================================

V3 版本有较大提升，基本解决了 V1 / V2 中存在的一些问题，美学和构图有较大改善。详见以下例图：

The version 3 has significant improvements, basically solving some of the issues present in V1/V2, with noticeable enhancements in aesthetics and composition. See the examples below:

1. 不同量化精度出图质量对比，量化损失很小。High-quality quantization parameters, the loss is very low.

   

2. 与官方及旧版模型对比效果。Comparison with the official and old versions.

   

3. V1 - V3 样例与对比。Examples and Comparisons of V1 - V3.

   

4. 仅提示词指令图像编辑。Just use instruct prompt to edit image.

   提示词：给他穿上粗犷的牛仔服外套，敞开着，露出他原来穿的T恤和图案，戴上一顶帽沿上翘的粗皮质牛仔帽，并让他嘴上叼个烟斗，烟斗上有一缕烟雾。保持图片的其它元素不变。

   Prompt: Put him in a rugged denim jacket, open to show the T-shirt and design he was originally wearing, add a thick leather cowboy hat with an upturned brim, and have him holding a smoking pipe with a wisp of smoke. Keep the other elements of the picture unchanged.

   

5. 仅提示词 + LoRA 换头、换装等。Prompt + LoRA to do Tryon / Tryoff and Faceswap.

   提示词：用图2的女士的脸，完全替换掉图1女士的脸。加 bfs_head_v1 LoRA 权重 0.75 (0.6-0.9)。

   Prompt: Use the face of the woman in picture 2 to completely replace the face of the woman in picture 1. (model + lora - bfs_head_v1 weight = 0.75)

   

6. 通过 Mask + LoRA 实现图像的区域精准编辑。Precise regional image editing using Mask + LoRA.

   (提示词和头像图片同上。The prompt and Face picture 2 are the same as above.)

   

使用方法, How to use：

与 V1 / V2 版本模型相同，The same with V1 / V2 vesrion.

或者参考 B 站如下大佬们的视频, Please refer to UP's video as below :

感谢：@T8star-Aix， @AI-KSK， @黑鹤001， @蓝色多脑盒， @初级AAA， @山竹老师AIGC， @炮老师的小课堂， @HyLan_L， @雪煮松风， @冰缘排骨， @暴躁小白兔学AI 以及 @活力网TV， 全平台统一号的 @CommintAI 等一众社区大佬对本模型的各种技术评测、详细使用介绍及进一步微调提升等方面的建议和支持。

Also on: Modelscope

特别致谢，Special thanks：

感谢：@亮亮rayne，@GuangyuanSD 在模型调制优化过程一如既往的给与技术、构图、美学、社群等方面的指导和支持。

Thanks to @亮亮rayne and @GuangyuanSD for their usual guidance and support in the model tuning and optimization process, covering technology, composition, aesthetics, community, and more.

=============================================================================

Flux2-Klein-9B-True-V2

***请不要下载该模型文件，烦请移步 HF 或 ModelScope 下载***

Due to network, Pls Download from : Huggingface.co, Modelscope.cn

V2 版本模型在 FLUX.2-Klein-9B-True-V1 的基础上再次进行了全量微调，相比 V1 模型，在如下方面得到了改善和提升：

1. 图像的真实感和质感进一步改善，基本接近香蕉(Nano Banana)的水平。

2. 提示词遵循和还原能力，参数适配性和LoRA兼容性进一步改善，。

3. 成图结构更归整、合理，美学方面也有一些改善。

4. 图像编辑能方面，提示词遵循更好，偏移和偏色也有一些改善。

具体效果参见如下对比图片。

The V2 version of this model has undergone a full fine-tuning based on FLUX.2-Klein-9B-True-V1. Compared to the V1 version, it has some improvements and enhancements as below:

1. The realism and texture of images have been further improved, basically reaching the level of Nano Banana.

2. Prompt adherence and restoration capability, parameter adaptability, and LoRA compatibility have been further enhanced.

3. The image structure is more organized and reasonable, with some aesthetic improvements as well.

4. In image editing ability, prompt adherence is better, pixel shift and color bias have also seen some improvements.

For specific results, please refer to the comparison images as below:

使用方法：

The same with V1 vesrion.

for Turbo Lora : anyMODE's rank128 / rank64 Civitai or Huggingface, weight 0.2 - 0.3 for steps 4 - 8.

for Distill LoRA : vafipas663's rank128 / rank256 Huggingface, weight 0.4 - 0.6 for steps: 4 - 8.

for Color and Shape correct of Image, you can use plugin of : ComfyUI-LCS

=============================================================================

Flux2-Klein-9B-True-V1

本模型为 FLUX.2-klein-9B 的微调模型，相比官方模型，去蒸馏、更清晰、更写实，模型编辑能力更精准，大大减少蒸馏模型过低步数带来的细节崩坏问题。具体效果参见如下示例图片，本模型易于使用、参数适配性和LoRA兼容性良好，如果结合 Real-Qwen-Image-V2 模型的中文字体渲染和多风格图像生成能力，AI生图创作如虎添翼。

This model is a fine-tuned version of FLUX.2-klein-9B. Compared to the official model, it is undistilled, clearer, and more realistic, with more precise editing capabilities, greatly reducing the problem of detail collapse caused by insufficient steps in distilled models. For specific effects, see the example images below. This model is easy to use, has good parameter adaptability, and is compatible with LoRA. If you combined with the Real-Qwen-Image-V2 model's Chinese font rendering and multi-style image generation capabilities, The image creation will greatly enhanced.

使用方法：

basic workflow pls refer to: image

inference cfg: 1.0；

inference steps: 20 -30；

edit or inpainting steps: 10-25;

or add turbo lora about steps: 8-12;

sampler / scheduler: euler / simple ，or any other。

Also on: Huggingface.co, Modelscope.cn

单区域编辑示例：

多区域+参考图编辑示例：

​

附加提示：

Turbo Lora Link：anyMODE's rank128/rank64 Klein Turbo Lora。

推荐使用 wuli大雄oO 大佬的 Flux2 Klein 超强多功能工作流，功能齐全又方便。

模型许可：

本模型请在遵循 black-forest-labs 公司非商业许可 (FLUX Non-Commercial License) 下使用。

This model falls under the FLUX Non-Commercial License.