🎨 Imagined by Rose — GPU Examples

Model file: flux-2-klein-4b.safetensors

Variant: Reference composition

Run route: Local GPU · stable-diffusion.cpp native API

Generation time: Not recorded in the saved receipt.

Requested output: 1536 × 320 · Steps: 28 · Seed: not recorded

Job ID: job_6ad4c594_00000000

Prompt

Use the attached live KingDing.co.uk Rose & Etta banner as the direct composition and pixel-art style reference. Create a new but unmistakably matching extra-wide website banner: exact same visual grammar and character roles, with Rose (pink-haired girl in burgundy witch hat and cape) and Etta (blonde girl in cream and sage-green clothing) standing together with the small white cat in the lower-left; keep them small and fully visible, grouped at far left beside a simple rose-emblem sign. A broad sunlit cobblestone garden path leads through the open center toward distant soft castle cliffs beneath a peach-pink and lavender evening sky. A cozy timber garden shop with warm square windows and climbing roses occupies the right side. Dense dark green foliage and rose beds frame the lower edge and corners. Match the reference's crisp 16-bit pixel RPG illustration, restrained pixel clusters, same warm palette and readable silhouettes; not glossy, not painterly, not anime. Keep the existing balanced proportions: characters small, shop moderate in size, path dominant. The banner must crop cleanly to 1485x300. Remove all words and lettering from signs and shop; no logo, UI, frame, watermark, extra characters, or extra pets. This is a website concept for review only, not an approved brand asset.

Negative prompt

photorealism, 3D render, painterly brushwork, watercolor, anime close-up, oversized faces, oversized characters, characters cropped, characters seated, extra people, extra animals, dogs, robots, noisy microdetail, excessive roses, distorted hands, text, letters, numbers, fake typography, watermark, user interface, border, logo

Model file: flux-2-klein-4b.safetensors

Variant: Composition refinement

Run route: Local GPU · stable-diffusion.cpp native API

Generation time: Not recorded in the saved receipt.

Requested output: 1536 × 320 · Steps: 20 · Seed: not recorded

Job ID: job_6ad4c6a1_00000001

Prompt

Transform the supplied live banner image into a fresh Rose & Etta website-banner concept while preserving its composition and original 16-bit pixel-art style. Keep the same scene structure and relative scale as the source: Rose and Etta small and fully visible at far left with a single small white cat beside them; simple wooden rose signpost behind them; broad path crossing the center; distant castle on the left-center horizon; warm wooden shop with lit windows on the right. Keep the girls at about 40 percent of image height, not enlarged. Preserve their established colors and roles: Rose pink-haired in burgundy witch hat and cape, Etta blonde in sage green and cream. Make the one cat anatomically clear and separate from the girls; no duplicate animals. Keep the open center bright with warm rose-pink sunset sky, soft green hedges and a few red roses. Match the source's crisp, chunky, limited-palette pixel clusters and slightly textured RPG landscape. Remove all existing text: turn every wooden sign and shop sign into blank wood with a small rose symbol only. Do not add any lettering. No extra people, no extra pets, no large foreground characters, no UI, no logo, no border, no watermark. New website concept only, not approved as an official brand asset.

Negative prompt

extra cats, duplicated cats, fused animals, distorted cat, extra characters, oversized characters, close-up, people on ledges, photorealism, glossy 3D, painterly brush strokes, anime face close-up, unreadable text, typography, lettering, numbers, logos, watermark, interface, frame

Model file: flux-2-klein-4b.safetensors

Variant: Signage cleanup

Run route: Local GPU · stable-diffusion.cpp native API

Generation time: Not recorded in the saved receipt.

Requested output: 1536 × 320 · Steps: 20 · Seed: not recorded

Job ID: job_6ad4c6f8_00000002

Prompt

Make a new panoramic Rose & Etta garden banner using the supplied live KingDing banner as style and composition reference. Keep the same crisp 16-bit pixel-art world, warm sunset palette, modest storybook scale and panoramic view. Rose and Etta remain small full-body figures together at far left in their established outfits, with exactly one small white cat on the path beside them. Preserve the rose garden path through the center toward distant castle cliffs, and the warmly lit timber flower shop on the right. The girls should be about 40 percent of canvas height, not enlarged. Change the foreground and shop details slightly, but keep the original left-to-right layout and path as the dominant central shape. IMPORTANT: remove all signboards, plaques, banners and shop signs completely; fill those areas naturally with green leaves, rose bushes, lanterns or timber. No text-like marks or lettering anywhere. Keep one single white cat with a clear natural body and tail; no duplicates. Match the live banner's pixel scale, edges, limited palette and gentle warm light. No extra people or animals, no close-ups, no UI, no logo, no border, no watermark. Unapproved website concept.

Negative prompt

signboards, signs, plaques, banners, text, letters, misspelled words, numbers, logo, watermark, extra cats, duplicate animals, fused animals, extra characters, close-up, oversized people, photorealistic, painterly, 3D render, anime, interface, border

Model file: flux-2-klein-4b.safetensors

Variant: Character identity

Run route: Local GPU · stable-diffusion.cpp native API

Generation time: Not recorded in the saved receipt.

Requested output: 1536 × 320 · Steps: 18 · Seed: not recorded

Job ID: job_6ad4c752_00000003

Prompt

Create a new website banner variation from the supplied KingDing Rose & Etta reference. Preserve its pixel-art style and exact left-to-right composition. Character identity is essential: Rose has vivid pink hair, a pointed burgundy witch hat and burgundy cape over cream clothes; Etta has pale blonde hair and a sage-green/brown adventurer outfit. Keep them as the same two small full-body companions grouped at the far left, not new characters and not close-ups. One small white cat stands clearly beside their feet. Keep their original hair colors, costumes and relative order. The cobbled path must dominate the center and lead toward the distant castle; keep the cozy timber shop only in the far-right quarter, with rose vines and warm window light. Keep all signboards removed and do not render any text or lettering. Maintain the reference's chunky pixel clusters, sunset rose-gold sky, dark green foliage and red roses. Avoid changing the camera framing. No extra people, animals, signage, logos, UI, borders or watermark. Unapproved website concept.

Negative prompt

photorealistic, painterly, blurry, smooth gradients, anime close-up, oversized characters, text, lettering, watermark, interface, border, logo

Model file: flux-2-klein-4b.safetensors

Variant: Signage cleanup

Run route: Local GPU · stable-diffusion.cpp native API

Generation time: Not recorded in the saved receipt.

Requested output: 1536 × 320 · Steps: 18 · Seed: not recorded

Job ID: job_6ad4c79d_00000004

Prompt

Generate a fresh extra-wide 16:5 Rose & Etta website banner. Use the supplied live KingDing.co.uk image as style reference only; do not copy its written signs. Compose a crisp, polished 16-bit pixel-art scene like the site: small full-body Rose and Etta together in the lower-left fifth with one small white cat at their feet; central open cobblestone path through a rose garden leading toward distant castle cliffs under a peach-pink sunset; compact warmly lit timber shop on the far right. Character identity must match the reference: Rose has vivid pink hair, pointed burgundy witch hat and burgundy cape; Etta has blonde hair and sage-green/cream adventurer clothes. Keep both girls smaller than half the banner height; no close-up. Use clear readable silhouettes, restrained chunky pixel clusters and the same warm rose-red, leafy green, amber, cream and lavender palette. Include lush roses but keep the center path bright and uncluttered. Do not include any signboards, plaques, banners, lettering, words, symbols, logos, UI, border or watermark. Exactly two girls and exactly one white cat. Website concept only; not an approved asset.

Negative prompt

signs, signs with text, signboards, placards, text, letters, numbers, pseudo-writing, logos, watermark, UI, border, extra people, extra animals, duplicate cat, malformed cat, cropped character, giant character, portrait close-up, photorealistic, 3D, painterly, watercolor, soft blurry pixel art, excessive micro-detail

Model file: flux-2-klein-4b.safetensors

Variant: Framing and scale

Run route: Local GPU · stable-diffusion.cpp native API

Generation time: Not recorded in the saved receipt.

Requested output: 1536 × 320 · Steps: 16 · Seed: not recorded

Job ID: job_6ad4c842_00000005

Prompt

Create a new ultra-wide 16-bit pixel-art banner for KingDing.co.uk, visually consistent with the supplied Rose & Etta banner reference. Match its panoramic camera and environment scale. Rose and Etta should be small full-body game characters, not foreground portrait figures: together they occupy only the leftmost 15 percent of banner width and each is no taller than 38 percent of image height. Rose is pink-haired with a burgundy pointed witch hat and cape; Etta is blonde in sage green and cream. Put exactly one small white cat beside their feet, clearly separate and with a normal single tail. Leave the middle wide and open: a sunlit cobblestone path through rose hedges toward distant castle cliffs and a peach-pink sky. Put a compact wooden shop with warm windows in only the far-right quarter. Keep lush but restrained red roses framing corners and bottom, not covering the path. Match the reference's crisp pixel clusters, warm amber light, dark leafy greens, rose reds, cream and lavender-pink sunset, with coherent 16-bit RPG detail. Remove all signboards, plaques and lettering. No text, no symbols, no logos, no UI, no border, no watermark, no extra people or pets. Website concept only.

Negative prompt

photorealistic, painterly, blurry, smooth gradients, anime close-up, oversized characters, text, lettering, watermark, interface, border, logo

Model file: flux-2-klein-4b.safetensors

Variant: Lighting and palette

Run route: Local GPU · stable-diffusion.cpp native API

Generation time: Not recorded in the saved receipt.

Requested output: 1536 × 320 · Steps: 16 · Seed: not recorded

Job ID: job_6ad4c8d9_00000006

Prompt

Create a fresh 16:5 panoramic KingDing Rose & Etta website banner. The supplied live banner is a close visual style and color reference: bright, warm, crisp 16-bit RPG pixel art with a rose-pink and apricot sunset, lush but light readable garden, clear small character silhouettes. Rose (pink-haired, pointed burgundy witch hat and cape) and Etta (blonde, sage-green and cream outfit) stand together at far left with exactly one small white cat. Keep the character group close to the reference scale, around half the banner height, fully visible. Let a wide bright cobblestone path lead through the open center to a distant castle. A compact warm timber flower shop sits on the far right, no more than one quarter of width. Use medium leafy greens and warm golden highlights; avoid near-black foliage and large shadow masses. A few roses frame the edges, keeping the center open. No signs or signboards anywhere; no words, letters, symbols, logo, UI, frame, watermark, extra people or animals. Maintain the source image's pixel clustering and palette; do not switch to glossy anime or painterly rendering. Website concept only, not an approved asset.

Negative prompt

dark black-green foreground, huge shop, signage, signboards, text, letters, pseudo-writing, logos, watermark, UI, frame, extra people, extra cats, duplicate cat, malformed animal, close-up, cropped figures, photorealism, glossy 3D, watercolor, painterly brushwork, blurry

Model file: flux-2-klein-4b.safetensors

Variant: Palette and contrast

Run route: Local GPU · stable-diffusion.cpp native API

Generation time: Not recorded in the saved receipt.

Requested output: 1536 × 320 · Steps: 16 · Seed: not recorded

Job ID: job_6ad4c973_00000007

Prompt

Panoramic 16:5 pixel-art website banner concept for KingDing.co.uk. Use the live Rose & Etta header as a precise style reference. Keep medium, natural contrast like the reference: vivid rose-red flowers and warm amber shop windows against leafy mid-to-dark greens and a pink/apricot sunset. Do not wash the scene out with pale haze, bloom or overexposure. Keep sky and path bright but preserve rich colors and crisp dark outlines. In the far-left lower corner, Rose and Etta are small full-body characters around 40 percent of banner height, with one small white cat at their feet. Rose has pink hair, burgundy pointed witch hat and cape; Etta has blonde hair and sage-green/cream clothes. Keep the girls grouped at the left and the rest of the banner mostly environment. A wide cobbled garden path fills the central foreground and leads toward distant castle cliffs. Compact lit timber flower shop at far right, no more than one quarter of width. A few red roses frame the corners, center stays open. Match the source's clean 16-bit RPG pixels and deliberate clustered shading; not painterly, not glossy anime. No signboards, no signage, no text, no symbols, no lettering, no logo, no UI, no border, no watermark, no extra characters or animals. Website concept only.

Negative prompt

photorealistic, painterly, blurry, smooth gradients, anime close-up, oversized characters, text, lettering, watermark, interface, border, logo

Model file: flux-2-klein-4b.safetensors

Variant: Shop details

Run route: Local GPU · stable-diffusion.cpp native API

Generation time: Not recorded in the saved receipt.

Requested output: 1536 × 320 · Steps: 16 · Seed: not recorded

Job ID: job_6ad4ca17_00000008

Prompt

Generate a panoramic KingDing.co.uk Rose & Etta banner concept in the same crisp, polished 16-bit pixel-art style and color balance as the supplied live image. Preserve the visual proportions of the reference: Rose and Etta, in established pink-haired witch and blonde sage-green adventurer outfits, are small full-body figures grouped at the lower-left with one small white cat. The girls are around half the banner height and the party remains at the edge, not in close-up. A wide bright cobbled path carries the eye through the center to distant fairytale cliffs and castle. A compact warm timber shop occupies only the far-right quarter; emphasize its roofline, glowing square windows, lanterns, climbing roses and leafy eaves. Do not put any signs, boards, plaques, banners, ribbons, labels or lettering on or near the shop or anywhere else; replace all such elements with plain wood, roof tiles and leaves. No text-like marks at all. Keep a rose-pink and apricot sunset, mid-tone leafy greens, saturated red roses and amber light, matching the live reference without washed-out haze or near-black shadows. Exactly two girls and one cat. No logo, UI, frame, watermark, extra character or pet. Website concept only.

Negative prompt

photorealistic, painterly, blurry, smooth gradients, anime close-up, oversized characters, text, lettering, watermark, interface, border, logo

Model file: flux-2-klein-4b.safetensors

Variant: Wide framing

Run route: Local GPU · stable-diffusion.cpp native API

Generation time: Not recorded in the saved receipt.

Requested output: 1536 × 320 · Steps: 16 · Seed: not recorded

Job ID: job_6ad4caad_00000009

Prompt

Final candidate in a ten-image GPU review set: create an extra-wide panoramic 16-bit pixel-art banner for KingDing.co.uk, closely matched to the live Rose & Etta header's establishing-shot composition. Pull the camera back so the garden environment dominates. The two full-body girls plus exactly one small white cat occupy only the lower-left 14 percent of image width; each girl should be about one-third of the banner height, small compared with the castle, hedges and shop. Rose: pink hair, burgundy pointed witch hat and cape. Etta: blonde hair, sage-green and cream adventurer outfit. Keep them fully visible beside one another, single cat clearly visible at their feet. The center half of the banner is a wide open cobbled rose-garden path leading to distant castle cliffs under a vivid pink, apricot and lavender sunset. The timber flower shop should occupy no more than the far-right 22 percent, with amber square windows and rose vines, not a close-up. Mid-tone leafy green framing, scattered saturated red roses, bright but not washed out. Crisp readable 16-bit pixel clusters, same palette, density and warm storybook style as the provided live reference. No foreground close-ups, no signs or boards, no words, no lettering, no symbols, no logo, no UI, no border, no watermark, no extra people or animals. Website concept only, not approved as a brand asset.

Negative prompt

close-up, foreground-sized characters, giant shop, oversized building, signboards, signage, text, writing, letters, logos, watermark, UI, border, extra characters, extra cats, duplicate cat, malformed cat, photorealism, glossy 3D, painterly, watercolor, washed-out haze, black shadows

Model file: flux-2-klein-4b.safetensors

Variant: Character scale

Run route: Local GPU · stable-diffusion.cpp native API

Generation time: Not recorded in the saved receipt.

Requested output: 1536 × 320 · Steps: 18 · Seed: not recorded

Job ID: job_6ad4cb64_0000000a

Prompt

Refine this existing GPU-generated Rose & Etta banner concept for closer fit to the live KingDing.co.uk header. Preserve its successful crisp 16-bit pixel art, warm pink/apricot sunset, castle path, rose garden, wooden shop, and exact character identities. Keep exactly two girls: Rose with pink hair and burgundy pointed witch hat/cape; Etta with blonde hair and sage-green outfit. Keep exactly one small white cat. Do not add or duplicate people or animals. Make the two girls about 10 percent smaller and keep them fully visible at the lower left so the environment remains dominant. Keep the shop at the far right, slightly more compact, with warm windows and climbing roses. Keep the central path open and bright. Do not add any signboards, plaques, banners or text. No lettering, symbols, logo, interface, border or watermark. Maintain the same palette and pixel scale as the input; do not overexpose or intensify the magenta sky. Website concept only.

Negative prompt

photorealistic, painterly, blurry, smooth gradients, anime close-up, oversized characters, text, lettering, watermark, interface, border, logo

Model file: flux-2-klein-4b.safetensors

Variant: Character visibility

Run route: Local GPU · stable-diffusion.cpp native API

Generation time: Not recorded in the saved receipt.

Requested output: 1536 × 320 · Steps: 16 · Seed: not recorded

Job ID: job_6ad4cbb0_0000000b

Prompt

Keep this GPU-generated Rose & Etta banner almost exactly as it is: same pixel-art style, two girls, character outfits, garden, path, castle, timber shop, warm light, framing and colors. Make only one small change: add exactly one small, clearly visible white cat on the ground beside the girls' feet at the far left. The cat must be separate from the clothing and path, with one normal body, four legs, two ears and one tail. Do not add any other animals or people. Do not alter Rose's pink hair and burgundy witch outfit or Etta's blonde hair and sage outfit. Keep all signs and lettering absent. No words, logos, UI, frame or watermark. Website concept only.

Negative prompt

photorealistic, painterly, blurry, smooth gradients, anime close-up, oversized characters, text, lettering, watermark, interface, border, logo

Model file: flux-2-klein-4b.safetensors

Variant: Character scale

Run route: Local GPU · stable-diffusion.cpp native API

Generation time: Not recorded in the saved receipt.

Requested output: 1536 × 320 · Steps: 16 · Seed: not recorded

Job ID: job_6ad4cbf5_0000000c

Prompt

Make one restrained refinement to this GPU-generated Rose & Etta banner while preserving the overall composition, pixel-art rendering, sunset colors, garden path, castle, timber shop, roses, and lighting. Keep exactly the same two girls and one white cat, with the same outfits and identities. Reduce the girls and cat together by about 10 percent and place them naturally at the far-left ground edge, making them feel slightly farther into the landscape. Keep their full bodies and the single white cat visible. Do not add, duplicate, remove or change characters. Keep the right-side shop and open center path as they are. No signs, no text, no logos, no UI, no border, no watermark. Preserve crisp 16-bit pixels and the source image's moderate color contrast.

Negative prompt

photorealistic, painterly, blurry, smooth gradients, anime close-up, oversized characters, text, lettering, watermark, interface, border, logo

Model: Flux-2-Klein-4B

Variant: Original benchmark settings

Generation time: 31.404 seconds

Model load time: 4.82 s

Hardware: AMD Radeon RX 9060 XT · 16188 MB dedicated VRAM

Resolution: 1024 × 1024 · Steps: 4 · CFG: 1.0 · Seed: 420013

Original image: 01_Flux-2-Klein-4B.png

SHA-256: 4cf0bce7d923b44145049822da55ff2afe2fdea5303b60acd010c2b53f1ef827

Prompt

Premium cinematic editorial illustration for a comedy platform. A charming expressive AI dog sits at a beautifully cluttered wooden library table beside a friendly retro-futurist robot, both earnestly preparing for a serious book discussion. Between them sits one absurdly polite turnip treated like a respected third member of the study group. Rich tactile detail: realistic fur, brushed metal, old book cloth, paper fibers, warm wood grain. Sophisticated visual storytelling, subtle British observational-comedy tone, believable physical scene, natural gestures, premium magazine illustration quality, cinematic depth of field, soft window light mixed with warm library lamps, restrained deep ink, cobalt and lilac accents, nuanced colour grading, excellent composition and focal hierarchy. No text, no speech bubbles, no logos, no watermark, no fake UI. Setup only; do not reveal or reference the potato punchline.

Model: Flux-2-Klein-Base-4B-Q8

Variant: Original benchmark settings

Generation time: 246.567 seconds

Model load time: 0.00 s

Hardware: AMD Radeon RX 9060 XT · 16188 MB dedicated VRAM

Resolution: 1024 × 1024 · Steps: 50 · CFG: 4.0 · Seed: 420013

Original image: 03_Flux-2-Klein-Base-4B-Q8.png

SHA-256: 3a0b82284868d06c81a0c8cd3aecc699128ae932fa10cbfe2319be2f258d6567

Prompt

Premium cinematic editorial illustration for a comedy platform. A charming expressive AI dog sits at a beautifully cluttered wooden library table beside a friendly retro-futurist robot, both earnestly preparing for a serious book discussion. Between them sits one absurdly polite turnip treated like a respected third member of the study group. Rich tactile detail: realistic fur, brushed metal, old book cloth, paper fibers, warm wood grain. Sophisticated visual storytelling, subtle British observational-comedy tone, believable physical scene, natural gestures, premium magazine illustration quality, cinematic depth of field, soft window light mixed with warm library lamps, restrained deep ink, cobalt and lilac accents, nuanced colour grading, excellent composition and focal hierarchy. No text, no speech bubbles, no logos, no watermark, no fake UI. Setup only; do not reveal or reference the potato punchline.

Model: SD-Turbo

Variant: Original benchmark settings

Generation time: 3.318 seconds

Model load time: 5.58 s

Hardware: AMD Radeon RX 9060 XT · 16188 MB dedicated VRAM

Resolution: 512 × 512 · Steps: 4 · CFG: 1.0 · Seed: 420013

Original image: 06_SD-Turbo.png

SHA-256: c852465bd317c863e7619f0d7f73ec4bc2cadf33af8fe8d84ea7c8ccf3c79563

Prompt

Premium cinematic editorial illustration for a comedy platform. A charming expressive AI dog sits at a beautifully cluttered wooden library table beside a friendly retro-futurist robot, both earnestly preparing for a serious book discussion. Between them sits one absurdly polite turnip treated like a respected third member of the study group. Rich tactile detail: realistic fur, brushed metal, old book cloth, paper fibers, warm wood grain. Sophisticated visual storytelling, subtle British observational-comedy tone, believable physical scene, natural gestures, premium magazine illustration quality, cinematic depth of field, soft window light mixed with warm library lamps, restrained deep ink, cobalt and lilac accents, nuanced colour grading, excellent composition and focal hierarchy. No text, no speech bubbles, no logos, no watermark, no fake UI. Setup only; do not reveal or reference the potato punchline.

Model: SDXL-Base-1.0

Variant: Original benchmark settings

Generation time: 34.077 seconds

Model load time: 4.63 s

Hardware: AMD Radeon RX 9060 XT · 16188 MB dedicated VRAM

Resolution: 1024 × 1024 · Steps: 20 · CFG: 7.5 · Seed: 420013

Original image: 07_SDXL-Base-1_0.png

SHA-256: deee20c927e9b483980f91e20ade527c395f31f1c4f564d40db270ddb2ab9e06

Prompt

Premium cinematic editorial illustration for a comedy platform. A charming expressive AI dog sits at a beautifully cluttered wooden library table beside a friendly retro-futurist robot, both earnestly preparing for a serious book discussion. Between them sits one absurdly polite turnip treated like a respected third member of the study group. Rich tactile detail: realistic fur, brushed metal, old book cloth, paper fibers, warm wood grain. Sophisticated visual storytelling, subtle British observational-comedy tone, believable physical scene, natural gestures, premium magazine illustration quality, cinematic depth of field, soft window light mixed with warm library lamps, restrained deep ink, cobalt and lilac accents, nuanced colour grading, excellent composition and focal hierarchy. No text, no speech bubbles, no logos, no watermark, no fake UI. Setup only; do not reveal or reference the potato punchline.

Model: SDXL-Turbo

Variant: Original benchmark settings

Generation time: 12.728 seconds

Model load time: 3.97 s

Hardware: AMD Radeon RX 9060 XT · 16188 MB dedicated VRAM

Resolution: 1024 × 1024 · Steps: 4 · CFG: 1.0 · Seed: 420013

Original image: 08_SDXL-Turbo.png

SHA-256: 09f4778dfd0ff258b7768fae91857be2ea4dafde35054269a9868a9a632250a2

Prompt

Premium cinematic editorial illustration for a comedy platform. A charming expressive AI dog sits at a beautifully cluttered wooden library table beside a friendly retro-futurist robot, both earnestly preparing for a serious book discussion. Between them sits one absurdly polite turnip treated like a respected third member of the study group. Rich tactile detail: realistic fur, brushed metal, old book cloth, paper fibers, warm wood grain. Sophisticated visual storytelling, subtle British observational-comedy tone, believable physical scene, natural gestures, premium magazine illustration quality, cinematic depth of field, soft window light mixed with warm library lamps, restrained deep ink, cobalt and lilac accents, nuanced colour grading, excellent composition and focal hierarchy. No text, no speech bubbles, no logos, no watermark, no fake UI. Setup only; do not reveal or reference the potato punchline.

Model: SD-Turbo-GGUF

Variant: Original benchmark settings

Generation time: 2.677 seconds

Model load time: 11.86 s

Hardware: AMD Radeon RX 9060 XT · 16188 MB dedicated VRAM

Resolution: 512 × 512 · Steps: 4 · CFG: 1.0 · Seed: 420013

Original image: EXP_SD-Turbo-GGUF.png

SHA-256: 5345b131b62f6d83bbf7ef59159739325c52fbb78541cd08b22cac9d2cda7b38

Prompt

Premium cinematic editorial illustration for a comedy platform. A charming expressive AI dog sits at a beautifully cluttered wooden library table beside a friendly retro-futurist robot, both earnestly preparing for a serious book discussion. Between them sits one absurdly polite turnip treated like a respected third member of the study group. Rich tactile detail: realistic fur, brushed metal, old book cloth, paper fibers, warm wood grain. Sophisticated visual storytelling, subtle British observational-comedy tone, believable physical scene, natural gestures, premium magazine illustration quality, cinematic depth of field, soft window light mixed with warm library lamps, restrained deep ink, cobalt and lilac accents, nuanced colour grading, excellent composition and focal hierarchy. No text, no speech bubbles, no logos, no watermark, no fake UI. Setup only; do not reveal or reference the potato punchline.

Model: SD-1.5

Variant: Original benchmark settings

Generation time: 11.917 seconds

Model load time: 10.68 s

Hardware: AMD Radeon RX 9060 XT · 16188 MB dedicated VRAM

Resolution: 512 × 512 · Steps: 25 · CFG: 7.0 · Seed: 420013

Original image: EXP_SD-1_5.png

SHA-256: 01a2f2adf5fbbad58754d1ed11c982406e8dac250b38234210a7da9cfa7fcbc3

Prompt

Premium cinematic editorial illustration for a comedy platform. A charming expressive AI dog sits at a beautifully cluttered wooden library table beside a friendly retro-futurist robot, both earnestly preparing for a serious book discussion. Between them sits one absurdly polite turnip treated like a respected third member of the study group. Rich tactile detail: realistic fur, brushed metal, old book cloth, paper fibers, warm wood grain. Sophisticated visual storytelling, subtle British observational-comedy tone, believable physical scene, natural gestures, premium magazine illustration quality, cinematic depth of field, soft window light mixed with warm library lamps, restrained deep ink, cobalt and lilac accents, nuanced colour grading, excellent composition and focal hierarchy. No text, no speech bubbles, no logos, no watermark, no fake UI. Setup only; do not reveal or reference the potato punchline.