anime2real-semi is a LoRA that converts anime-style images into a semi-photorealistic / realistic CG style.
During training, proportions were strictly controlled, so the resulting body shape and face structure reference the original anime features, including NSFW content. That's why I call her "real-semi". Feel free to try her out!
LoRA weight: 0.9
Base model: Klein 9b
It's extremely fun and I highly recommend it to you!
Workflow: Klein 9b anime2real-semi
Experience link: https://www.runninghub.ai/post/2069000025157496833/?inviteCode=368034ba
anime2real-semi 是一个将动漫风格图片转换为半写实 / 写实CG风格的 LoRA。
训练时严格控制了比例，转换出来身形、脸型、都会参考动漫原有的特征，包括NSFW等。 因此我称她为“real-semi”。欢迎大家尝试。
Description
Version：β
FAQ
Comments (19)
好用至极，特来夸奖，几乎是我用过最好的转cos写实的lora了
道友慧眼如炬，我这属于半写实
Damn, you're not joking about the nsfw parts. This is REALLY GOOD!
this lora is amazing but can you do another version which will not make face realistic and will retain shapes of the drawing but in 3d semirealism? kinda like vinyl figures or 3d renders but not only anime style .most of the dataset was asian girls and it shows , is it possible to disable face realism ?
Hahahaha, I'll adjust it later when I have time.
Straight up amazing!
Thank you! Really glad you like it
works well but seems to neutralize most of face reactions
Yes, I only trained it on a very small dataset, so it wasn’t optimized for facial expressions.
I created a workflow that compared the lora effects of multiple anime-to-real human conversions, including this lora（anime2real-semi，The NSFW effect is quite good.）.You can click the link to give it a try.
Thanks to the author for providing such a powerful tool for our community!
4b version planned?
works very well.
maybe anyone know a way to add a little realism or color saturation?
Yes, you can improve realism and color by adding prompt terms like HDR, natural lighting, film grain, or vibrant colors. It can help enhance saturation and overall detail a bit.
My only issue with the lora is that the images it outputs are very washed out. If it was better at retaining the input images colors it would be perfect.
You can try adding "HDR" in your prompt for a moderate improvement in visual results. I’m also collecting higher-quality datasets myself.
整体脸部太幼态了 有些丝袜和贴身的衣服都好像识别不太出来，暗肤色也是 希望出个V2优化一下
这个lora对原图的脸的比例有一定要求，越接近正常比例效果越好，如果之后有好的训练集会考虑再训练
Details
Files
anime2real-semi.safetensors
Mirrors
anime2real-semi.safetensors
anime2real-semi.safetensors
anime2real-semi.safetensors
anime2real-semi.safetensors
anime2real-semi.safetensors
anime2real-semi.safetensors
anime2real-semi.safetensors
anime2real-semi-mid_2341496-vid_2633750.safetensors
Anime2real-semi - Klein9B - 转为写实风格,Convert to photorealistic style,realistic,turn the image into a realistic photo.safetensors
Anime2real-semi - Klein9B - 转为写实风格,Convert to photorealistic style,realistic,turn the image into a realistic photo.safetensors
anime2real-semi.safetensors
anime2real-semi-mid_2341496-vid_2633750.safetensors
24-anime2real-semi.safetensors
anime2real-semi.safetensors
anime.safetensors
anime.safetensors
anime.safetensors