Join The Tinkerer on Whop to download this model and many more — all for one monthly payment. Plus, get early releases, private tools and a lot more. 👉 Join on Whop
CyberRealistic Z-Image Turbo is a build based on the original Z-Image Turbo. It follows the original Z-Image Turbo settings and philosophy, with minimal intervention, exploring how well this setup translates into the CyberRealistic workflow.
⚙️ Personal Settings (Forge Neo)
or
Required Additional Files
Make sure you also have the following:
16 GB+ VRAM: qwen3_4b.safetensors
8–12 GB VRAM: qwen34bfp8_scaled.safetensors
All VRAM sizes VAE: ae.safetensors
This model is shared as-is, for users who enjoy experimenting, benchmarking, and pushing models outside the comfort zone.
Feedback, findings, and edge cases are welcome - this release exists primarily to learn from real-world usage.
Description
Yes, it jumps from v1.0 to v1.3.
No, it’s not a revolution.
Just a cleaner, tighter, slightly more caffeinated version of the original.
Same soul. Fewer hiccups. More “oh damn, that looks good.”
FAQ
Comments (24)
I know that this question about the forge was asked before; I was just wondering since it was mentioned that this can work with some modification in the standard, non-neo version of the forge. May I please ask what modifications are required to get this amazing Zimage turbo model working in the standard classic forge? Any help or advice is greatly appreciated, and thank you for creating such amazing models. They are truly the best I have ever seen!
Me too
Whynot upgrade to NeoForge? It's an upgraded and better version of the original Forge
Required Additional Files
Make sure you also have the following:
16 GB+ VRAM: qwen3_4b.safetensors
8–12 GB VRAM: qwen34bfp8_scaled.safetensors
All VRAM sizes VAE: ae.safetensors
⬆️ where to put this files? like which one where
qwen3_4b and qwen34bfp8 are text encoders and go into your webui/models/text_encoder
ae is a VAE and goes into your webui/models/vae
@1015 Thx mate
Great work again! Tho, i was wondering what the "RF" in DPM++ 2s a is.
It's something "exclusive" from Forge Neo
Коллеги , я смотрю тут интересные темы начали подниматься..
Вопрос у меня к вам : кто нибудь умудрился запустить z-image на старой GTX карте типа 1080Ti. Что-то у меня не особо получается...
новые версии portable ComfyUI ставится не хотят. Говорят: либо ошибка c10.dll, либо поставьте новый видео-драйвер. А он уже итак стоит. Новый Forge-Neo тоже вылетает с ошибкой, хотя версия чуть старее работает отлично, но блин она не-видит z-image .
Поставить новый ComfyUI через Git не удается из за тормозов с инетом, грустно...
Testing with the help of Grok
The GTX 1080 Ti is based on the Pascal architecture and has Compute Capability 6.1 7.
It was officially supported by CUDA up to version 12.x, but starting with CUDA 13, NVIDIA dropped support for the Pascal architecture (including the GTX 1080 Ti) 2.
However, users are successfully using:
CUDA 11.x (e.g., 11.3)
CUDA 12.6 with PyTorch 2.9
Use CUDA 11.8 or 12.1—the most stable for Pascal.
Try the latest comfyui build with CUDA 12.6 on GitHub
ComfyUI_windows_portable_nvidia_cu126.7z
If that doesn't work, downgrade your PyTorch version: Install a version built with CUDA 11.8/12.1 support, for example:
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
Запускал на GTX 1070 и Qwen image и Qwen edit 2509 ComfyUI Устанавливал с помощью чат ботов, GPT, Grok, ComfyUI_portable версию. Чат боты в помощь.
Проблема решена,
качаем новую версию comfyUI portable ,
качаем и устанавливаем VisualCppRedist_AIO_x86_x64 2017-2026.exe
сносим python если стоит в системе ( в архиве все равно свой есть)
чистим реестр от старых записей
распаковываем ComfyUI запускаем, материмся на новый интерфейс.
закидываем файлики по свом местам и вуаля после перезагрузки все заработало...
Разве что с какими-нибудь шакальными гуфами, иначе это буде ооочень долго.
@Cyberdelia I understand it is experimental and you want feedback.
It would be great if you could include data you have used in CR Pony v11 specifically because that one was unique and really useful. But sadly for some reason you took out those elements in later releases, it seems.
hi, dude! How about kissing?
Did you know that the original z-image-turbo can't do a passionate (French kiss)?
Really? That doesn't even make sense... I feel like ZIT may not be all it's hyped up to be if it can't even understand daily basic things (and French kissing isn't really NSFW per se) without it being manually trained in.
I'd love to see a CyberRealistic Classic ZIT version
I like this, and some areas are improvements, but there's clearly some flux type synthetic images in there, that make it less realistic.
What do you prefer for realism then?
@ygmdir Well I don't know every z-image model (there's quite a few, and it's early days), but I've found 0.5 better for this in cyber realistic.
nsfw?
Can we get a Lora Version please! 😍
I found something interesting. Since Qwen has really good response to Chinese prompts, I translate some English prompts into Chinese by using google translate. For comparison, the English prompts always land in some wrong areas and dissatisfied contents, but the Chinese prompts can finish the jobs perfectly (even if I don't recognize any Chinese words).
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.