Anime2Real is a Qwen-Edit Lora designed to convert anime characters into realistic styles. The current version is beta, with characters appearing somewhat greasy. The Lora strength must be set to <1.
You may click the link below to test the LoRa effect: Anime2Real
V1.0 Update: This version makes characters appear less greasy and significantly more natural. The default generation produces Western facial features; to generate Asian or other facial types, simply load an additional Realistic Series LORA.
Description
V1.0 Update:
This version renders characters with a less greasy appearance and significantly greater naturalism. The default output features Western facial features; to generate Asian or other ethnicities, simply load an additional Real Series LORA.
FAQ
Comments (26)
我试了好像没有效果，只能变成2.5d，不知道是不是lora没生效
你的图能发出来看看吗.我试了很多,除了3D类型的几乎都能转
Does version 1.0 have a trigger phrase?
As with version 0.9, it converts images into a realistic style.
@shadowsii Yes, but version 0.9 had a trigger phrase: "将图片转为真实风格"
Version 1.0 doesn't list any, so I'm asking if it has no trigger phrase.
@Jellai My apologies, that was an oversight on my part. What I meant to say was that the trigger words for version 1.0 are identical to those in previous versions.
The restoration of colors, costumes, and prop details is excellent, but the ability to grasp the characteristics of appearance is slightly inferior, or the "cosplay" flavor is too strong.
This LoRa model was trained using only a small training dataset, resulting in limited generalisation capabilities. It may be beneficial to combine it with https://civitai.com/models/2121900.
Please mix the training dataset with non-asian people too. For some reason most of the anime2real models make every character look asian, even changing the eye color and hair color. Thank you!
My LORA presets all feature Western facial features. When testing on Asian subjects, I've found it necessary to layer an additional LORA. Should you require this effect, simply incorporate a photorealistic LORA.
not working. even after fully updating comfyui all nodes and dependencies and installing missing nodes. I still get error that i am missing these nodes:
NunchakuQwenImageDiTLoader
NunchakuQwenImageLoraLoader
fast prompts
fast concatPrompt
BizyAirSiliconCloudVLMAPI
You may directly utilise the native qwen to load nodes and replace the missing ones; the VLM is not essential.
The prompt word nodes in the workflow can't be found at all. Could you make some changes？
Perhaps you're referring to this:
大模型也需要写实模型吗，光是lora的话完全无效
不用啊,这是edit模型
要求大模型为qwen-image-edit
In your workflow you use lora Azer-QWEN写实 / Azer-qwen4654.safetensors . Please share it or give a link to it .
This is merely a photorealistic model designed to enhance Asian facial features. You may replace it with any photorealistic LROA model; it is not essential.
Is this model censored?? or Qwen is censoring NSFW ?
not nsfw
Does not seem to be working on Image-Edit-2511 version. At least, I tried, but no results.
I have retrained it for version 2511. Please visit https://civitai.com/models/2287977
@shadowsii Thanks, will test it when i get home!
Input: 😏😃😫😮😠
Output: 😐😐😐😐😐
This is just a fishing LoRA.It's a decoy.
Details
Files
Anime2Real_v4-22.safetensors
Mirrors
Anime2Real_v4-22.safetensors
Anime2Real_v4-22.safetensors
Anime2Real_v4-22.safetensors
Anime2Real_v4-22.safetensors
Anime2Real_v4-22.safetensors
Anime2Real_v4-22-mid_2110229-vid_2396073.safetensors
Anime2Real_v4-22.safetensors
qwen_Anime2Real_将图片转为真实风格_v4-22.safetensors
Anime2Real_v4-22-mid_2110229-vid_2396073.safetensors