Tsunade

Qwen-image:

Qwen-image is really good at anime and 3D! It won't be long before someone starts fine-tuning it, possibly in a month, two months, and we might see some super amazing checkpoints popping up.

1. It can be precise and true to the original character.

2. Good results can be achieved with minimal training—it’s easy to train.

3. It far surpasses models like SDXL, Flux, and Wan in terms of , image quality, text generation, and the resolution of generated images.

Wan2.1_T2V_14B_V1:

etc.

Hunyuan_V1:

I rarely use high resolution when generating with HunyuanVideo. It's not like the Lora quality is bad – it's more that cranking up the resolution eats up way too much time and gets expensive fast. If that's not a concern for you, though, go ahead and try running these Lora at 720P resolution. No doubt, the results will be even better.

NoobAI(ill) V-Pred:

It's basically just the same old training dataset anyway, all for the sake of Rule34.

Notes:

1. This model requires V-Pred support. SwarmUI and Comfy are recommended.

2. It's best to include 'blacked-out region', 'Masked Area', and tags in your negative prompts.

I thought I had already gotten rid of SDXL, but the community and the sheer power of are just too strong, relentlessly breathing life into SDXL. it, I really wish there were a better alternative.

Checkpoint: https://civitai.com/models/833294/noobai-xl-nai-xl

Flux-V1:

Cut some parts

Clothing: Only keep the green jacket and gray short-sleeved shirt,

Face shape: middle-aged face, young face,

Hairstyle: double braids, single ponytail

Pony-V1

Key promprt:

tsunade,Hokage’s hat

tsunade's Default Dressing \(clothes\),

White crop top,red skirt,fishnet,

tsunade's blue jacket \(clothes\),

tsunade's green Jacket \(clothes\),

tsunade's sandals \(sandals\),