cappuccinoimage

cappuccinoimage is a HiDream-O1-Image-based, supertuned MLX text-to-image checkpoint for local image generation. It is not an official HiDream.ai release. The upstream HiDream-O1-Image code and model weights are licensed under the MIT License.

Quick Start

Download the repository, then point a compatible local MLX generation runner at the downloaded model directory.

python path/to/generate.py \
  --model-path /path/to/cappuccinoimage \
  --model-type full \
  --scheduler unipc \
  --num-inference-steps 50 \
  --width 2048 \
  --height 2048 \
  --seed 4101 \
  --prompt-file prompt.txt \
  --output output.png

The examples below use full, UniPC, guidance_scale=5.0, and shift=3.0.

Examples

Example 1

• Prompt: Editorial interior photograph of a cozy winter living room, mustard-yellow sofa, brick fireplace, traditional framed artwork, textured armchair, layered rugs, soft amber window light, realistic materials, carefully balanced composition.

• Size: 2304x1728
• Steps: 50
• Seed: 4101

Example 2

• Prompt: High-end studio flat lay of a blue frosted celebration cake with rainbow sprinkles and a single candle, pastel pink background, clean product-photography lighting, crisp frosting texture, playful color contrast.

• Size: 2048x2048
• Steps: 50
• Seed: 4102

Example 3

• Prompt: Gallery-quality abstract watercolor painting with translucent red, cobalt blue, and warm orange swirls, delicate paper grain, controlled pigment blooms, elegant negative space, modern fine-art composition.

• Size: 2048x2048
• Steps: 50
• Seed: 4103

Example 4

• Prompt: Black-and-white cinematic film noir street scene, sharply dressed man standing under rain-slick city lights, glowing neon sign in the background, dramatic shadows, misty atmosphere, classic 35mm still-frame look.

• Size: 2304x1728
• Steps: 50
• Seed: 4104

Example 5

• Prompt: Wide cinematic science-fiction landscape with glowing alien plants, crystalline ground, a lone explorer in a spacesuit, distant futuristic city, purple twilight sky, strong depth, luminous environmental detail.

• Size: 2560x1440
• Steps: 50
• Seed: 4105

Step Comparison

Same prompt, size, seed, scheduler, guidance, and shift. Only the step count changes.

Example 6

• Prompt: Anime-inspired fashion portrait in a neon Las Vegas casino, androgynous model in a tailored white rabbit-themed jacket, playing-card accessories, glossy magazine lighting, vibrant nightlife background, polished character design.

• Size: 1440x2560
• Steps: 30
• Seed: 4201

Example 7

• Prompt: Anime-inspired fashion portrait in a neon Las Vegas casino, androgynous model in a tailored white rabbit-themed jacket, playing-card accessories, glossy magazine lighting, vibrant nightlife background, polished character design.

• Size: 1440x2560
• Steps: 50
• Seed: 4201

Example 8

• Prompt: Anime-inspired fashion portrait in a neon Las Vegas casino, androgynous model in a tailored white rabbit-themed jacket, playing-card accessories, glossy magazine lighting, vibrant nightlife background, polished character design.

• Size: 1440x2560
• Steps: 75
• Seed: 4201

Seed Comparison

Same prompt, size, scheduler, guidance, shift, and step count. Only the seed changes.

Example 9

• Prompt: Anime-inspired fashion portrait in a neon Las Vegas casino, androgynous model in a tailored white rabbit-themed jacket, playing-card accessories, glossy magazine lighting, vibrant nightlife background, polished character design.

• Size: 1440x2560
• Steps: 50
• Seed: 4202

More Examples

Example 10

• Prompt: High-resolution landscape photograph of a turquoise alpine lake framed by tall pine trees and snow-capped mountains, clear sunny sky, crisp reflections, natural color grading, expansive travel-photography composition.

• Size: 2560x1440
• Steps: 50
• Seed: 4210

Example 11

• Prompt: Editorial fashion photograph of a short-haired model wearing a flowing emerald-green dress outdoors under a bright blue sky, clean styling, natural movement in the fabric, refined magazine lighting.

• Size: 1440x2560
• Steps: 50
• Seed: 4211

Example 12

• Prompt: Minimalist studio product photograph of a matte red-and-orange ceramic bottle with intricate surface patterning, placed on a curved pedestal, warm directional lighting, soft shadows, premium design-catalog styling.

• Size: 2048x2048
• Steps: 50
• Seed: 4212

Example 13

• Prompt: Fine-art profile portrait of a woman with flowing blonde hair dissolving into a starry galaxy, gold light particles woven through the hair, deep blue circular backdrop, ethereal atmosphere, elegant photographic detail.

• Size: 1440x2560
• Steps: 50
• Seed: 4213

Example 14

• Prompt: Cinematic cyberpunk fashion portrait of a pale woman with blue eyes, short asymmetrical hair, bold black eyeliner, neon red city lights, high-detail styling, shallow depth of field, dramatic night atmosphere.

• Size: 1440x2560
• Steps: 50
• Seed: 4214

Example 15

• Prompt: Soft high-key anime-inspired portrait with a clean white theme, delicate layered clothing, gentle expression, bright studio lighting, subtle fabric texture, polished character illustration, calm minimalist composition.

• Size: 1440x2560
• Steps: 200
• Seed: 4215

Example 16

• Prompt: Candid editorial portrait of a young woman relaxing in a leather cinema seat, tied white shirt, wide-leg jeans, gold necklace, long dark hair, soft screen-glow lighting, shallow-depth 85mm fashion-photography look.

• Size: 1440x2560
• Steps: 50
• Seed: 4216

Prompt Agent Rewrite Comparison

The prompt agent uses cappuccinodense to expand a short input prompt before image generation.

Same size, seed, step count, scheduler, guidance, and shift. The second image uses the prompt rewritten at example-generation time.

Example 17 - Input Prompt

• Prompt: cat on the desk

• Size: 2048x2048
• Steps: 50
• Seed: 4301

Example 18 - Prompt Agent Rewrite

• Prompt: A photorealistic close-up of a fluffy orange tabby cat curled up asleep on a wooden desk, bathed in warm afternoon sunlight streaming through a nearby window. The cat’s soft fur is detailed with natural texture, its paws tucked under its body, eyes closed peacefully. The desk surface shows subtle wood grain, with a few scattered items: a ceramic coffee mug, a leather-bound notebook, and a pair of reading glasses. Shallow depth of field keeps the cat in sharp focus while the background softly blurs into a cozy home office setting. Shot on 85mm lens, natural lighting, cinematic composition, warm color grading, high detail, 4K quality.

• Source prompt: cat on the desk
• Rewritten with: cappuccinodense
• Size: 2048x2048
• Steps: 50
• Seed: 4301