Mysterious - Qwen Update
Please kindly rate my model with 5 stars and leave the images you've created with it. I would greatly appreciate that and it would bring me much happiness.
Alternatively, you can support me with a cup of coffee.
https://www.buymeacoffee.com/LahIntheFutureland
Thank you to everyone who has supported the Mysterious Series so far.
After the earlier releases on SDXL and Flux, this new version takes a significant step forward — now upgraded to Qwen Image.
What’s Improved
This edition focuses on enhancing both artistic fidelity and pose stability, while preserving the signature identity of the Mysterious style.
Higher image clarity and richer textures
Much stronger performance on complex poses, including:
dynamic / expressive action poses
multi-layer composition and atmospheric depth
cinematic camera angles
Preserves the iconic Mysterious aesthetic:
soft golden-hour lighting and pastel tone
gentle rimlight and airy fantasy ambience
a blend of Eastern mysticism with modern fantasy flavor
The goal of this upgrade is not just to generate “beautiful images” —
but to expand the creative range, allowing scenes that feel emotional, narrative, and alive.
About Us — LahGenAI
Our team includes known creators in the AI art and workflow community such as: @Lah_Inthe_Futureland , @DucHaiten, @Hius, ChugAx from shakker and @Hideroki
Visit us at: https://lahteam.vn
Our Services
We specialize in:
Finetuning Checkpoints and LoRAs for image and video applications.
Building ComfyUI workflows for image/video processing and integrating them into AI production pipelines.
Deploying and packaging AI products, including solution consulting and product development support for startups.
If you need to build a stable, scalable visual AI pipeline, we can support your team end-to-end.
Thank you again for supporting the Mysterious Series.
Feel free to share your creations — we truly love seeing what you build. ✨🎴
Discord: douglastran
Description
FAQ
Comments (6)
I see that your finetuning has very good results, can I ask how long did you train on what GPU
A100 from google colab
@LahIntheFutureland Thank you, I'm okay with asking how many steps you've FT ？ has your loss dropped significantly? My loss has been around 0.1 after smoothing and has not changed
@HyperToto I'm not entirely certain about the training settings, as each time I train, I split the data for processing multiple times, and each time I set different parameters based on the generate img when testing versions. The number of training steps depends on whether the model has learned from the data I've provided or not. If it hasn't learned yet, I'll continue training for a few more steps based on my assessment.
Personalmente, me encantó, no lo inclinaría a ninguna cultura en especial, dale la fantasía como prioridad y no necesariamente que sea ilustración, de manera realista da buenos resultados, generando un image to image con Photon o Realistic Vision.
Thanks! Love u so much! <3