We all know how difficult and frustrating it can be to generate several characters simultaneously in one picutre, especially with Illustrious and Pony models.It is particularly difficult if the characters must be generated with their own specific LoRA.I recently explored a few techniques and worked a few ComfyUI workflows to try to solve this issue.This work and the associated workflow are avaliable in my article here:<h3 id="how-i-generate-pictures-with-several-characters.-d8vf39mhf"><a rel="ugc" href="https://civitai.com/articles/15657">How I generate pictures with several characters</a>.</h3>Now, we want to go further: the goal of this bounty will be to improve these techniques, discover new ones and, this is important, share them with the community.Your challenge - if you accept it - will be to generate images with several specified characters without prompt bleeding.Your submissions:- will have to show at least two different characters,- all characters must be specifically distincts, either through specific prompts or through specific LoRA,- the more characters the better! - you can use any model and any LoRa of your choice,- any characters, any style are ok,- SFW and NSFW are ok, just please, refrain yourself from posting hardcore porn or XXX pictures.This is the perfect occasion to feature your original characters or reuse existing ones into dynamic and beautiful compositions!I will award two separate trophies:<ul><li>10 000 Buzz for the most beautiful picture.</li></ul>Absolutely no pre-conditions here (except the ones above). Creativity, style, characters interactions and composition will be key here.<ul><li>10 000 Buzz for the best generation technique. </li></ul>Our goal here is to learn new ways to generate several characters in a picture, so this is the most interesting trophy. So, obviously, all generation data including workflow(s) and tools will have to be shared with the picture.Simplicity of use, coherence of image/characters and reliability in the generation are what we are looking for here.You can get bonus points by using my LoRA (<a rel="ugc" href="https://civitai.com/models/1528502/clara-by-coloneljay">Clara</a> and the <a rel="ugc" href="https://civitai.com/models/1717041?modelVersionId=1943212">Medieval chaperon</a>) and posting your submissions on the LoRA page. This is completely optional but, many thanks in advance if you choose to use them.Feel free to set your characters into mini-stories or sceneries.There will be additionnal Buzz tips awarded depending on the quality of your submissions.It is possible for a submission to win both prizes but I guess I wish to reward more people so probably there will be two distinct winners here.I set the max submissions limit to 50, which I suppose is largely enough for most of you.The most important thing here: Lets'shave fun, get creative and unleash your skills: we need more multi-characters pictures here on Civitai!

img_00024final_.png

img_00170uf2_.png

img_00003_.png

img_00048_.png

Multi-characters images generation - 20k Buzz rewards!

Submitted Images: ComfyUI_00472_.png (1.51 MB) - "The Guardian's Roar" - Fierce red dragon defending treasure hoard with smaller companion dragons - Multiple distinct dragon characters with clear scale and color differentiation ComfyUI_00469_.png (1.57 MB) - "Ancient Keeper" - Majestic earth-toned dragon with human-scale figures for perspective - Demonstrates size hierarchy between massive dragon and tiny human characters ComfyUI_00505_.png (1.53 MB) - "The Sage Dragon" - Wise dragon in classical architecture with multiple human characters - Shows architectural scale with dragon and several distinct human figures ComfyUI_00506_.png (1.56 MB) - "Mother &amp; Hatchlings" - Purple-scaled dragon with distinctive offspring showing clear character variation - Parent-child dragon relationship with different scale colors and sizes ComfyUI_00510_.png (1.45 MB) - "Crystal Cave Encounter" - Turquoise dragon with adventurers in mystical cavern setting - Multiple character types (dragon + human adventurers) in complex environment Bounty Submission Categories:"Most Beautiful Picture" - Submitting all 5 images for artistic merit"Best Generation Technique" - Pure Pony Diffusion anti-bleeding method Multi-Character Achievement: - 2-4+ distinct characters per image  - No prompt bleeding - Clean character separation  - Specific character traits - Unique colors, sizes, roles  - Diverse character types - Dragons, humans, various scales  Technical Innovation - Generation Technique: Methodology: "Pure Pony Diffusion Multi-Character with Anti-Bleeding Negatives" Model Stack: - Base: ponyDiffusionV6XL_v6StartWithThisOne.safetensors (pure model, no LoRA) - Sampler: DPM++ 2M, Simple scheduler, 100 steps, CFG 6.9 Critical Anti-Bleeding Technique: Negative Prompt: score_5, score_4, score_3, score_2, score_1, zPDXL3, bad anatomy, bad hands, merged characters, prompt bleeding Multi-Character Strategy: 1. Character Count Declaration - "multiple characters" at prompt start 2. Scale Hierarchy - "titanic dragon" vs human-scale characters 3. Color/Type Differentiation - Explicit descriptions per character 4. Role-Based Separation - Each character has distinct actions/positionsGeneration Data for Submitted Images: - Files: ComfyUI_00472_.png, ComfyUI_00469_.png, ComfyUI_00505_.png, ComfyUI_00506_.png, ComfyUI_00510_.png - Model: ponyDiffusionV6XL_v6StartWithThisOne.safetensors - Resolution: 1024x1024, Steps: 100, CFG: 6.9 - Scheduler: Simple, Sampler: DPM++ 2M - Workflow: f3d1.json (included)