About the Creator
Toki_no | AI Model Creator | Free Workflow Sharing
π Click the link to test this model online! New users get 1000 points + 100 daily bonus, enough for ~60 images!
If you'd like to test the model's performance, you can register using the link below. I've set up a workflow there, where you can earn free credits to try it out and then decide whether to download the model!π Try it now: https://www.runninghub.ai/post/2015286346998353921/?inviteCode=rh-v1446
π Try it free online, no local setup needed! Like & Save if you enjoy!
QianShi-YingJian is a specialized LoRA built on the powerful Qwen-Image (Q-I) base model, fine-tuned specifically for
Asian Boudoir Photography (η§ζΏζε½±).
By leveraging Qwen-Image's superior semantic understanding, this model moves beyond standard SDXL limits. It captures
not just the pose, but the subtle intimacy and ambient dust of a private photo session with incredible fidelity.
π Why this Q-I Build?
* π§ Smarter Understanding: Thanks to the Qwen-Image base, it follows complex prompts much better than traditional
models.
* ποΈ Authentic Texture: optimized for skin texture and low-light environments (window light, lace shadows).
* π― Zero Struggle: No need for "magic tags" or "quality boosters". It just works.
---
βοΈ Photographer's Guide
ββββββββββββββββ¬βββββββββββββ¬βββββββββββββββββββββββββββββ
β Setting β Value β Notes β
ββββββββββββββββΌβββββββββββββΌβββββββββββββββββββββββββββββ€
β Base Model β Qwen-Image β Required for this LoRA. β
β Trigger Word β η§ζΏζε½± β The core trigger. β
β Weight β 0.8 β Recommended balance. β
β CFG β 3.5 - 5.0 β Keep lower for Q-I models. β
ββββββββββββββββ΄βββββββββββββ΄βββββββββββββββββββββββββββββ
π‘ Pro Prompt Formula
> η§ζΏζε½± + [Subject Description] + [Lighting/Mood]
Copy this Prompt:
> η§ζΏζε½±, a young woman sitting on a wooden chair, wearing a white slip dress, sunlight streaming through blinds,
casting shadows on her face, emotional eyes, film grain, raw photo, ultra realistic, 8k.
---
π Community & Resources
Join us to get the most out of this model:
* Online Demo: Try the Q-I workflow instantly.
* Free Points: Daily credits for generation.
* Support: Direct help from the creator.
π Click to Join
[Join Toki_no Community] (https://discord.gg/PeRQuxZm)
Description
FAQ
Comments (3)
Actual testing shows that Z-IMage supports LOA, and when used in a workflow, it only requires 8 steps to output.
Well, what is LOA then? (the technology and progress going so fast and I can not keep-up with everything or with every acronym)
Here's the situation: my description might be slightly inaccurate. The model was originally trained on Qwen, but I recently tested it and found that it also works correctly with ZImageTurbo. This might be because ZImageTurbo uses Qwen for more refined training.