Unified multimodal model for text-to-image generation, instruction-guided image editing, personalized generation, and virtual try-on
Settled by community votes across 13 shared challenges, with an AI judge weighing in on each.
OmniGen v2
16.9
arena score
#57 of 62 in Text-to-Image
Skill signature
· Text-to-Image
Stable Diffusion 3.5 Large Turbo
10.7
arena score
#61 of 62 in Text-to-Image
Vote tally
Where the votes landed
OmniGen v2
0%
win rate
Ties
0%
Stable Diffusion 3.5 Large Turbo
0%
win rate
Shared challenges
13
Challenge by challenge
The strongest take from each model on every shared challenge, with the AI judge's read.
Geometric Composition
Text-to-Image“A glass cube on a wooden table. Inside the cube is a small blue sphere. On top of the cube sits a red book. A green plant is behind the cube, partially visible through the glass. Soft window light from the left.”
OmniGen v2
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
OmniGen v2
- + Excellent spatial arrangement following all instructions.
- + High-quality realistic lighting and refraction effects.
- + Natural photorealistic style.
- − The plant behind the glass is quite blurry due to depth of field.
Stable Diffusion 3.5 Large Turbo
- + Sharp focus on the objects and the plant.
- + Crisp lighting and shadow work.
- − Failed spatial prompt: the red book is inside the cube rather than on top.
- − The glass cube has an unrealistic black frame not requested in the prompt.
- − The sphere is sitting on the book instead of the table.
Verdict: OmniGen v2 followed the spatial prompt perfectly, correctly placing the book on top of the cube and the sphere inside it. Stable Diffusion 3.5 Large Turbo failed to understand the relationship between the objects, placing the book and sphere inside a black-framed display case.
Candid Street Photography
Text-to-Image“A candid street photo of an elderly Japanese man repairing a red bicycle in light rain, reflections on wet pavement, shallow depth of field, 50mm lens, natural skin texture, imperfect framing, motion blur from passing cars, cinematic but realistic, no stylization.”
OmniGen v2
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
OmniGen v2
- + Excellent wet pavement reflections including the red bicycle
- + Accurately depicts the 'light rain' and environmental atmosphere
- + Strong composition with a believable depth of field and urban background
- − The man is posing with the bike rather than actively 'repairing' it
- − Lack of significant motion blur on the passing cars as requested
Stable Diffusion 3.5 Large Turbo
- + The subject's posture more closely resembles someone repairing a bicycle
- + Good skin texture on the man's hands and face
- − Significant anatomical and object errors including fused hands and a distorted bicycle frame
- − The rain effect is very sparse and lacks the requested environmental reflections
- − Background cars are static and lack any motion blur
Verdict: OmniGen v2 produced a much more coherent and aesthetically pleasing image with realistic lighting and reflections that captured the atmosphere of a rainy street. Stable Diffusion 3.5 Large Turbo struggled with the complex geometry of the bicycle and the man's hands, resulting in a distorted and less realistic image.
Fantasy Warrior
Text-to-Image“Close portrait of a battle-worn paladin in ornate engraved plate armor, hair braided with small beads, faint scars and dirt on the skin, warm torchlight reflecting off metal, shallow depth of field, bokeh sparks, lifelike eyes, highly detailed texture on leather straps and cloth underlayer.”
OmniGen v2
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
OmniGen v2
- + Excellent adherence to the 'beads in hair' and 'bokeh sparks' prompt details
- + The warm torchlight lighting is atmospheric and realistically reflects off the metal armor
- + High-quality skin texture and lifelike eye rendering
- − The character looks remarkably clean and polished for someone described as 'battle-worn'
Stable Diffusion 3.5 Large Turbo
- + Successfully captures a more rugged, battle-worn appearance with prominent scarring and grime
- + Ornate engraving on the armor is sharp and well-defined
- − The skin texture appears somewhat plastic or overly smoothed in several areas
- − Failed to include the requested beads within the braided hair
- − The lighting feels more like a studio light than natural torchlight
Verdict: OmniGen v2 produced a much more aesthetically pleasing and cohesive image that followed nearly every specific detail of the prompt, including the small beads and bokeh sparks. While Stable Diffusion 3.5 Large Turbo did a better job conveying the 'battle-worn' aspect with more severe scarring, the overall image quality and lighting in OmniGen v2 are superior.
Modern Clean Menu
Text-to-Image“Modern minimalist restaurant menu design, white background with colorful food photos in grid, sections for appetizers/pizza/mains, bold sans-serif fonts, vibrant accents, clean professional layout for casual dining.”
OmniGen v2
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
OmniGen v2
- + Excellent grid layout that looks like a real multi-page menu.
- + Captures the professional casual dining aesthetic with clear headings.
- + Strong use of vibrant accents and consistent design language.
- − Significant spelling errors throughout the headers.
- − The text within the menu lists is gibberish lines.
Stable Diffusion 3.5 Large Turbo
- + Excellent top-down food photography with vibrant colors.
- + Cleaner typography for 'Pizza' and 'Mians' headings.
- + Highly creative and appealing food presentation.
- − Layout feels more like a mood board or flyer than a functional restaurant menu.
- − Composition is a bit cluttered with items overlapping edges.
- − Missed the 'appetizers' section requested in the prompt.
Verdict: OmniGen v2 presents a far more realistic menu layout that adheres structurally to the request for sections and grids, though it struggles with exact spelling. Stable Diffusion 3.5 Large Turbo produces much higher quality food imagery and cleaner fonts, but the overall composition feels less like a professional document and more like a collection of assets.
Magic Burger Explosion: Fiery Photorealism Challenge
Text-to-Image“Ad for 'Magic Burger'. Dynamic, exploded burger with all components (bun, patty, cheese, lettuce, tomato, sauce) suspended in mid-air. Emphasize photorealistic detail and a sense of motion. Dark, fiery background with glowing embers. Integrate text: 'MAGIC BURGER' as a prominent title, 'LIMITED TIME ONLY' as a secondary message, and '€6.99' in a starburst, all rendered with a fiery, glowing effect.”
OmniGen v2
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
OmniGen v2
- + Successfully integrated all requested text with mostly correct spelling.
- + Followed the starburst price tag requirement.
- + Clean, professional graphic design layout suitable for an advertisement.
- − Failed to create an 'exploded' burger, showing a fully assembled one instead.
- − The text 'LIMITED TIME ONLY' is slightly cut off at the edge.
- − The lighting on the burger feels flat compared to the fiery background.
Stable Diffusion 3.5 Large Turbo
- + Better sense of motion with dripping sauce and glowing embers.
- + Dynamic lighting that interacts more realistically with the burger textures.
- + More creative interpretation of the fiery background with smoke effects.
- − Completely ignored all requested text elements.
- − Failed the 'exploded burger' requirement by keeping components stacked.
- − The bottom bun appears to be melting or morphing into liquid in a visually confusing way.
Verdict: OmniGen v2 is the clear winner for its superior prompt adherence, particularly regarding the complex text and layout requirements. While Stable Diffusion 3.5 Large Turbo captured the 'fiery' atmosphere and motion better, it failed to include any of the requested typography, which was a core part of the ad challenge.
Chalkboard Menu
Text-to-Image“Handwritten-style chalkboard menu in a cozy café, all text rendered in the exact same realistic chalk handwriting style with natural variations in letter size, slight slant, and chalk texture — no printed or digital fonts anywhere on the board. Title at the top in elegant cursive chalk handwriting: ‘TODAY’S SPECIALS – APRIL 30, 2026’. Below it, three menu items also in the same handwritten chalk style: ‘Truffle Mushroom Risotto – $24’, ‘Grilled Octopus with Lemon & Herbs – $28’, ‘Brown Butter Chocolate Chip Cookies – $9’. At the very bottom, smaller text in the identical handwritten chalk style (slightly smaller but still clearly legible with the same handwriting characteristics): ‘All items made fresh daily • Ask about our gluten-free options’. Warm ambient café lighting, visible chalk dust and smudges, realistic handwriting imperfections, no clean printed text anywhere.”
OmniGen v2
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
OmniGen v2
- + Successfully rendered the full date and prices correctly
- + Better texture on the chalk and chalkboard surface
- + Captures the handwritten variation in letter sizes well
- − Includes a spelling error in the title ('SPECALS')
- − Significant text garbling and repetition in the middle of the board
Stable Diffusion 3.5 Large Turbo
- + Excellent scene composition with plants and cafe lighting
- + The handwriting style is more elegant and follows a consistent aesthetic
- + Good spatial layout with three distinct columns
- − Failed to render specific prompts like the date and prices correctly
- − Many spelling errors in the menu items ('risctto', 'ocotpgs')
- − The text looks more like a digital font than actual chalk on a board
Verdict: OmniGen v2 followed the specific textual details and pricing much better than its competitor, though it suffered from severe logic issues and repetition in the central text area. Stable Diffusion 3.5 Large Turbo created a much more visually appealing and professional composition, but it failed significantly on the prompt adherence regarding the specific date, prices, and text legibility. OmniGen v2 is the narrow winner for actually including 'April 30, 2026' and the requested price points.
The Reversed Rodeo
Text-to-Image“Horse riding astronaut in space — horse on top, not vice versa. Surreal, highly detailed, cinematic.”
OmniGen v2
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
OmniGen v2
- + Clean, illustrative style with high-contrast colors.
- + Good detail on the astronaut's suit and the horse's mane.
- − Failed the negative constraint; the astronaut is riding the horse.
- − The anatomy of the horse's back legs is distorted and nonsensical.
- − Composition is static and lacks the cinematic feel requested.
Stable Diffusion 3.5 Large Turbo
- + Cinematic composition with a sense of motion and scale.
- + Lighting and shadows are well-integrated with the space background.
- + High level of detail on the space suit and Earth's surface.
- − Failed the spatial reasoning constraint; the horse is being ridden by the astronaut.
- − One of the horse's front legs has a deformed, split hoof.
- − The rider's hands are floating and not properly aligned with the reins.
Verdict: Both OmniGen v2 and Stable Diffusion 3.5 Large Turbo failed the negative constraint to have the horse riding the astronaut, providing the standard 'astronaut on a horse' image instead. Stable Diffusion 3.5 Large Turbo is the preferred image because its composition is more cinematic and dynamic compared to the flat, anatomically broken output from OmniGen v2.
The Capybara Taxi Driver
Text-to-Image“Photorealistic scene inside a yellow New York taxi at night. A capybara is driving, wearing a yellow taxi driver cap and a dark jacket. It has a calm, professional expression and both front paws on the steering wheel. In the back seat sits a human businesswoman in a coat, looking at her phone with a completely normal, bored expression (as if this is just another normal ride). Through the windows you can see the streets of Manhattan at night with blurred lights. Realistic taxi interior, photorealistic, detailed fur and fabric, 35mm lens, night lighting with reflections, shallow depth of field.”
OmniGen v2
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
OmniGen v2
- + Excellent character rendering of the capybara
- + Clear and detailed depiction of the businesswoman looking at her phone
- + Vibrant colors and sharp focus
- − The capybara has human hands instead of front paws on the steering wheel
- − The passenger is sitting in the front passenger seat rather than the back seat
Stable Diffusion 3.5 Large Turbo
- + Correctly depicts the capybara's paws on the steering wheel
- + Successfully places the passenger in the back seat as requested
- + Better captures the 'dark jacket' and night-time city lighting context
- − The passenger in the back is very blurry and lacks detail
- − The steering wheel placement looks slightly detached from the dashboard
Verdict: Stable Diffusion 3.5 Large Turbo followed the spatial instructions much better by placing the passenger in the back seat and correctly rendering the capybara's paws. While OmniGen v2 has higher facial clarity, it failed significantly on logical details by giving the animal human hands and putting the passenger in the front seat.
The Halloween Invitation
Text-to-Image“Vintage gothic Halloween party invitation. Dark parchment poster, spooky border with webs and thorns, central glowing jack-o-lantern, bats, twisted trees, moody night sky. Add elegant gothic title text saying "Halloween Party Invitation", a small scroll banner saying "You are invited to a night of frights", and event details at the bottom: Date: 30.10.2026 Time: 7pm Location: The Arches, NYC Spooky but polished, cinematic lighting, square format.”
OmniGen v2
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
OmniGen v2
- + Successfully included all segments of text including the specific date and location.
- + Excellent layout with a central glowing jack-o-lantern and thematic twisted trees.
- + Strong cinematic lighting that creates a genuine spooky atmosphere.
- − Several spelling errors in the smaller text fields (e.g., 'ARCRESOTED', 'FRIETS', 'ARCAS').
- − The banner text is jumbled and poorly rendered compared to the main title.
Stable Diffusion 3.5 Large Turbo
- + Clean and sharp graphic design style with a clear thorny border.
- + The central jack-o-lantern has high visual quality and depth.
- + Good use of space and composition for a square format.
- − Failed to include most of the required text, including the specific date, time, location, and the banner quote.
- − The overall image feels more like a template than a completed invitation.
- − Lacks the 'dark parchment' texture requested in the prompt.
Verdict: OmniGen v2 is the clear winner for its superior prompt adherence, successfully incorporating the complex text requirements and specific atmosphere details like the parchment background. While Stable Diffusion 3.5 Large Turbo produced a cleaner graphic, it failed to generate any of the specific event details (date, time, location), making it non-functional as an invitation based on the prompt.
Isometric Miniature Diorama Scenes
Text-to-Image“Create a clear, 45° top-down isometric miniature 3D cartoon scene of Japan's signature dish: sushi, with soft refined textures, realistic PBR materials, gentle lighting, on a small raised diorama base with minimal garnish and plate. Solid light blue background. At top-center: 'JAPAN' in large bold text, 'SUSHI' below it, small flag icon. Perfectly centered, ultra-clean, high-clarity, square format.”
OmniGen v2
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
OmniGen v2
- + Excellent text rendering with perfect spelling of 'JAPAN' and 'SUSHI'.
- + Accurate 45-degree isometric perspective on a clean diorama base.
- + High adherence to the solid light blue background and minimalist aesthetic.
- − The flag icon does not accurately represent the Japanese flag.
- − The sushi hybridizes nigiri and maki in a way that looks slightly anatomically incorrect for food.
Stable Diffusion 3.5 Large Turbo
- + Strong tactile textures on the rice and salmon roe toppings.
- + Good use of 3D depth for the signboard and miniature props.
- + Clean isometric layout with a believable wooden base texture.
- − Typos in the text, rendering 'SUSHI' as 'SIIHI'.
- − Failed the background requirement by uses a gradient/greyish blue instead of solid light blue.
- − The flags provided do not resemble the Japanese flag.
Verdict: OmniGen v2 is the clear winner because it followed the text instructions perfectly, rendering both 'JAPAN' and 'SUSHI' without errors, and strictly adhered to the solid light blue background request. While Stable Diffusion 3.5 Large Turbo had nice material textures, it failed on the primary text spelling and the background color requirement.
Adorable Baby Animals in Sunny Meadow
Text-to-Image“Hyper-photorealistic scene of fluffy baby animals—a golden retriever puppy, tabby kitten, baby bunny, and red fox kit—with big expressive eyes and ultra-detailed soft fur, playfully chasing butterflies and tumbling together in a lush wildflower meadow, warm golden sunrise light with god rays and dew sparkles, joyful wholesome vibe, 8K masterpiece.”
OmniGen v2
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
OmniGen v2
- + Successfully included multiple distinct animal types (dog, cat, fox/bunny hybrid).
- + Stronger adherence to the 'golden sunrise' and 'god rays' lighting request.
- + Vibrant and cheerful color palette that matches the 'joyful wholesome vibe'.
- − Artificial, 3D-render aesthetic rather than 'photorealistic'.
- − Missing the rabbit; the third animal appears to be a fox with bunny-like features.
- − Anatomical issues with the animals' paws and the butterflies' limb structures.
Stable Diffusion 3.5 Large Turbo
- + Excellent fur texture and back-lighting effects on the puppy.
- + Generally higher visual quality and cleaner edges.
- + Dynamic composition with animals shown in motion/tumbling as requested.
- − Failed to include the four requested animal types (missing fox and bunny).
- − The 'kitten' on the right appears more like a wild caracal or lynx hybrid.
- − Does not capture the 'warm golden sunrise' as well as the competitor.
Verdict: Both models failed to accurately provide all four specific animals requested in the prompt, especially the baby bunny and red fox. OmniGen v2 adhered better to the environmental lighting and overall atmosphere, while Stable Diffusion 3.5 Large Turbo produced higher quality fur textures and a more dynamic composition despite a more limited animal count.
Vintage Cafe Logo
Text-to-Image“Vintage minimalist restaurant logo for "Caffè Florian", retro cloche dome with steam and "Est. 1720" banner, classic typography, warm brown and cream tones, subtle texture on light background, vector emblem style.”
OmniGen v2
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
OmniGen v2
- + Clean vector emblem style that matches the minimalist request.
- + Accurate rendering of the year 'Est. 1720'.
- + Good balance and symmetrical composition.
- − Spelling error in the main text ('CAFFFLORIN' instead of 'Caffè Florian').
- − The cloche icon is extremely simplified, missing the handle detail.
Stable Diffusion 3.5 Large Turbo
- + Excellent texture and vintage aesthetic as requested.
- + Creative integration of the cloche dome with a coffee cup shape.
- + More sophisticated illustration style with nice shading.
- − Spelling error in the name ('Caffeé' rather than 'Caffè').
- − The banner overlaps the text in a messy way near the end of 'Florian'.
Verdict: Stable Diffusion 3.5 Large Turbo produced a much more visually interesting and 'vintage' textured result that creatively combined the cloche and coffee themes. While OmniGen v2 achieved a cleaner minimalist vector look, its failure to spell the primary brand name correctly and its overly simplistic icons make it less effective than the more artistic approach of Stable Diffusion.
Apollo 11: Journey to Tranquility
Text-to-Image“Create a clean, modern vector infographic poster about the Apollo 11 mission. NASA-inspired palette (navy, white, muted red, light gray). Flat-vector style, crisp lines, consistent iconography, subtle gradients only. Steps (stop at landing): 1. Launch (Saturn Vicon) 2. Earth Orbit (Earth + orbit ring icon) 3. Translunar (trajectory arc icon) 4. Lunar Orbit (Moon + orbit ring icon) 5. Descent (lunar module descending icon) 6. Landing (lunar module on the surface icon) Small supporting elements (minimal text): • Crew strip: three silhouette icons with only last names: Armstrong, Aldrin, Collins. • Landing site marker: Moon pin labeled "Tranquility" only. Layout constraints: generous margins, large readable labels, clean background with subtle stars. Vector-only, print-poster look, high resolution.”
OmniGen v2
Stable Diffusion 3.5 Large Turbo
AI Judge Analysis
OmniGen v2
- + Clean vector aesthetic that matches the requested flat-vector style.
- + Follows the color palette accurately with navy, white, and red.
- + Layout clearly differentiates the six requested steps in a grid.
- − Failed the primary subject by labeling it 'APOLO 17' instead of Apollo 11.
- − Text is heavily garbled with significant spelling errors throughout.
- − Iconography is generic and does not clearly represent specific objects like the Saturn V.
Stable Diffusion 3.5 Large Turbo
- + Strong composition with a narrative flow that captures the 'spirit' of the mission.
- + Included an astronaut and lunar landscape consistent with the landing theme.
- + Better typography layout, even if the words contain minor errors.
- − Missed the navy blue requirement in favor of a cream/teal vintage palette.
- − Did not follow the specific 6-step structure requested in the prompt.
- − Contains significant visual artifacts, such as the astronaut's legs merging with a ladder/structure.
Verdict: OmniGen v2 adhered better to the technical visual requirements, such as the color palette and the flat-vector style, but completely failed the primary subject title by labeling it 'Apolo 17'. Stable Diffusion 3.5 Large Turbo created a more visually interesting 'poster' with better text rendering (Apoll.o), but it ignored the specific 6-step infographic constraints and the requested NASA color scheme. OmniGen v2 is the likely winner for better following the structural and stylistic constraints of an infographic.
Next steps
Explore each model
Distilled version of SD 3.5 Large that generates high-quality images in just 4 steps, offering faster inference and reduced costs