🔥 Download IntoRealism SDXL ULTRA V1 on Ko-Fi via credit card or PayPal
OR
# 🔥 ONE MEMBERSHIP = FULL ACCESS
Your membership includes :
- models
- workflows
- scripts
- Exclusives
- future updates
👉 $25/month on Ko-Fi
📢 Join The Community
We're building a friendly space to chat, share creations, and get support.
👉 Click here to join the Discord
✨ IntoRealism Ultra is an experimental release that combines the strength of DMD2 with the finesse of a Light model, delivering a more realistic look than ever before.
👉 A rare feature : unlike most light models, negative prompts work perfectly, giving you much more control and precision over the final renders.
🔹 Key features:
DMD2 + Light blend
Enhanced realism (textures, skin, natural lighting)
Handles NSFW content very well
Fully compatible with negative prompts
Versatile: portraits, lifestyle, cinematic scenes
better LoRA adherence
⚠️ Note : this is an early test version. The model will be further optimized and refined in future updates for even greater realism and flexibility.
Usage
Sampling method : Euler A Karras or Exponential
Sampling steps : 6 - 12 (12 - 14 for V6)
Resolution : 896x1152 / 832x1216
CFG : 1.5 - 2
Workflow : https://civarchive.com/models/2383077/intorealism-ultra-workflow
Description
The V6 is definitely the most realistic version. They are all different, so this one leans towards realism.
I recommend more than 10 steps for a better result, between 10 and 14.
It is also possible to switch to 2 CFG.
FAQ
Comments (23)
this quickly turned into one of my favourites! Great detail, great prompt adherence, great skin detail. puts SDXL on another level IMHO
Not a complaint, but there just seems to be something different about the VAE that is baked into v1 and v2 compared with the VAE baked into v3-6 (v4 was somewhere in the middle). The later versions seem to apply a bluish ultra-dramatic look to the image.
Everything with the image generation looks great until the last second when the VAE is applied and it shifts to some dramatic style. I ran into this with your original IntoRealism models too.
I realize it could have something to do with my external gen workflow, but I get great results with other models and your early versions. Just trying to understand the VAE a bit more -- thanks for the great models.
hey. I don't seem to have that issue when generating. I use Forge on Pinokio since it gives the same results as ComfyUI. I have a 5070 Ti and run local. That seems like it may be a workflow issue. I hope you get it figured out.
If not, I recommend installing Forge through Pinokio. It's made my life much easier (especially with finding checkpoint paths from ComfyUI to Forge). I only use ComfyUI for creating music with Ace Step, generating videos with Wan 2.2, and cloning voices.
You can use with original sdxl vae, if its a thing.
Call me crazy but I am getting a sense of your IntoRealism Ultra Amateur with v1 and maybe v5 merged?
Whatever you did, I dig it. I really liked the prompt adherence with your amateur model and the lighting with texture of v5. Thanks for the wicked Thanksgiving gift, my friend! I hope yours was textbook! If you aren't American, then know I'll cheers to you, regardless!
This model is superb- what scheduler are you using? I am using Euler Ancestral w/ Karras scheduler and the results are some of the best I have seen, with my prompting being the limitation more than not.
Euler Ancestra / Karrras is the best in my opinion.
Is there any way to get a de-distilled version of this model for lora training purposes? I almost exclusively use this model and have found that training character have much better accuracy when trained directly on the used model then SDXL base. For example, my loras trained with Big Love look MUCH better than those on SDXL when used in Big Love. Since this model uses DMD2, I'm not too sure its possible to train loras on it without modification. I've tried, and they come out as a blurry mess.
I recommend creating Loras using the SDXL version of my model, which is the basis for the ULTRA versions.
@enzino If I do the training with the ultra model, can I use this distilled version as a generation model? What the user says is true, these versions are much better for realism than the ULTRA SDXL model.
Ты просто Бог)
This model does well following prompts
whats the schedual type please?
Проверил на веснушки. Практически, идеально) Спасибо, Коллега, за Ваш замечательный чекпоинт!
Даже без апскейлов. Спасибо за такую потрясающую Работу, Бро!!!
This model is amazing! Are there any chances new ones will come out in the future?
Thank you ! Yes, a V7 is planned before the end of the year.
@enzino Жлем с нетерпением, Маэстро!)
ZiT - это конечно здорово. Но SDXL - в умелых руках не уступает)
Вот щас смотрю на работы sdxl. Ну ладно там - силиконовые сиськи. Но зачем так наяривать по контрасту? Эстэтика что ли такая модная? Там один мой ген с собакой чего стоит. И это не хвастовство. С Рождеством, Коллеги! И с Наступающим!
Great model series, thanks for providing them :D
However, i noticed slight image corruption (green and purple lines on object edges) that seem to be created by the baked in VAE. When i use the default SDXL VAE, the image is fine. Tested with Forge and ComfyUI with the models "IntoRealism XL (SDXL V3)" and "IntoRealism Ultra V6.0".