For early access products and exclusive contents check: PolyPhaze | Generate reality @ PolyPhaze | Patreon
Unlock the next level of character design with this Premium, High-Quality LoRA model. Specially designed to maintain a character's core features across various dynamic or static poses, this model ensures consistency from all angles.
Whether you're creating for artistic projects or need reliable reference guides, this tool is perfect for generating characters that stay true to their original design, no matter the perspective.
Key Features:
Consistent character appearance across multiple views
Ideal for dynamic and static poses
Perfect for character creation and reference guides
Take your creations to new dimensions with this powerful and versatile tool!
Keep the prompt simple!
Example:
"This image is a CGI multiple view of a cybernetic man, wearing a red armor, futuristic style, in four different angles, front view, back view and side views. She's wearing a helmet suit, with neon visor shining.
The background is a futuristic environment, apparently a futuristic lab."
Description
FAQ
Comments (9)
This is amazing! Can we get one for 1.5 too?
I'm not sure if 1.5 resolutions supports the training, bcs the images used were pretty wide... Earlier, I checked for 1.5 models and found this one: multiple views - v1.0 | Stable Diffusion LoRA | Civitai
Prompt for different angles, generate an image much larger than high, and you'll get the same results with SD1.5 models. It has always been possible to do. Without any Lora.
Error message: The following keys have not been correctly be renamed: lora_unet_input_blocks_0_resnets_0_emb_layers_1.alpha, lora_unet_input_blocks_0_resnets_0_emb_layers_1.hada_w1_a, lora_unet_input_blocks_0_resnets_0_emb_layers_1.hada_w1_b, lora_unet_input_blocks_0_resnets_0_emb_layers_1.hada_w2_a, lora_unet_input_blocks_0_resnets_0_emb_layers_1.hada_w2_b,
List goes on. Think this one only works in automatic1111 and comfy?
running smoothly on Fooocus here...
@femmefatale_ same here...
Nice job!!
Great!!
The same can be done with SDXL vanilla, and could already be done in SD1.5, with a simple prompt (full body, different views, front, left, back, from above, ect) and a large image. The larger, the more angles/versions you get.
This is a nice Lora idea, sadly it doesn't do anything that's already possible. And worse, it tends to separate the different angles when you get a smooth transition of the background without any Lora.