High-Fidelity Anime-to-Real Dual Sampling Workflow

Explanation video:

Usage Instructions:

1.This workflow realizes the anime-to-real human conversion function through two generations with Qwen+Zimage. Qwen uses the LoRAs: anime2real2511 and anything2real_2601_A_final, while Zimage uses the LoRA: Kook_Zimage_RealisticFantasy_Turbo. Images generated by Qwen retain the original content but have an overly strong 3D effect. The secondary generation with Zimage achieves a realistic photo effect and greatly enhances the human-like realism.

2.Multiple models are provided. The QwenEdit2511 + beyondREALITY combination can meet most requirements in general. For special needs, you can choose AIO23 + moodyPornMix_zitV7.

3.An automatic prompt generation function is available. You can select the model resolution according to specific needs, or manually reverse-engineer prompts via Gemini and Gork for manual input.

Model Download Links:

AIO23：https://huggingface.co/Phr00t/Qwen-Image-Edit-Rapid-AIO

moodyPornMix：https://civitai.com/models/620406/moody-porn-mix

anime2real2511：https://civitai.com/models/2287977/anime2real-2511

anything2real_2601_A_final：https://civitai.com/models/2121900?modelVersionId=2571125

Kook_Zimage_真实幻想_Turbo：https://modelscope.cn/models/KookYan/Kook_Qwen_JIZHIZHENSHI/files?version=20250905153225

This workflow is available for free online testing; download it for local use if you find it useful:

Online Workflow: https://www.runninghub.ai/post/2015763807888478210/?inviteCode=rh-v1443

Get 1000 points upon registration and an extra 100 points for daily login. Each image takes about 4 minutes to process and costs around 60 points.