Overview
There are many possible paths to get from a character idea to consistent and controllable character renders because everyone has different ways of thinking and working, unique starting assets and output requirements.
So here's a curated selection of several ways to get from A to B, designed to be as flexible as the creative process.
SECTION 1
Character design
Depending on what kind of ideas, sketches or images you're starting with, there are different workflows to turn those into final character images that you're happy with and ready to use for training.
Start with a written description, a sketch or an existing image for starters. Add in a face of someone if there's a particular person you have in mind.
Start with a written description, a sketch or an existing image for starters. Add in a face of someone if there's a particular person you have in mind.
Text2Image + Prompt Enhancer LLM for Flux
Flux Text to Character Sheet
Image to Character Spin
SECTION 2
Training Set Creation & LoRA Training
Once you have an image of a character that you like, it's time to create a variety of images that are based on that initial image. A character sheet is a good start although we recommend additional variations that have different environments and lighting, so that the LoRA doesn't associate the character with image properties and elements that will change on and around them.
In general, the more robust the variety of inputs, the more flexible and controllable the outputs will be while retaining the true essence of what the character is and isn't. Don't worry if there's minor inconsistencies in the input images, this is why we're creating a model in the first place!
In general, the more robust the variety of inputs, the more flexible and controllable the outputs will be while retaining the true essence of what the character is and isn't. Don't worry if there's minor inconsistencies in the input images, this is why we're creating a model in the first place!
Flux Kontext - Single Image to Character LoRA
ComfyUI Flux LoRA Trainer
SECTION 3
Image Creation & Manipulation
Once your LoRA is created, there are many ways to generate consistent images from that model, and manipulate it to match your ideas and requirements.
Flux ControlNet 2.0 - All-in-one
Flux Kontext Inpainting
Image to 3D with Hunyuan3D w/ Texture Upscale
SECTION 4
Video & Animating
Bring your assets to life. These workflows add motion with precisely the level of control you want, ranging from a single text prompt to detailed image, video, and audio references.
Wan2.1 FusionX Image2Video
Start/End Frame Multi-Video via Floyo API
SECTION 5
Audio, LipSync and VFX
Elevate your footage with cinematic polish. These workflows layer in rich audio, precise lip-sync, dynamic effects, adaptive camera moves, smart masking, outpainting, and more: everything you need to edit or enhance existing scenes.
Wan2.1 and FantasyTalking - Image2Video Lipsync