Updated: Sep 5, 2025
posesVerified: 3 days ago
SafeTensor
The FLUX.1 [dev] Model is licensed by Black Forest Labs. Inc. under the FLUX.1 [dev] Non-Commercial License. Copyright Black Forest Labs. Inc.
IN NO EVENT SHALL BLACK FOREST LABS, INC. BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.
Born out of frustration of how nano-banana and kontext max handled asymmetrical characters when putting them in an "a-pose", I went ahead and trained a Kontext LoRA that can bring characters from multiple different poses to an a-pose. This was trained from actual 3D renderings all from mihoyo on blender with dataset consisting of multiple characters each having multiple different unique poses taken from frames of Mikumikudance motions and the default a-pose. From testing, this seems to handle asymmetrical character's clothing colors much better than nano-banana.
Note: The character that nano-banana performed bad on was the asymmetrically colored summer outfit yuzuha \(zenless zone zero\) in which even fed front facing image, ended up making both sides mirrored instead of asymmetrical as designed.
This LoRA is not meant to compete with the RefControl LoRA as this one was trained on real blender renderings instead of people. This can in fact complement that LoRA by being a preprocessing step to that LoRA