candy - CivArchive (CivitAI Archive)

candy - v1.0

NSFW

Candy — Photorealistic Portrait Base Model (SD 1.5)

Candy is a fine-tuned Stable Diffusion 1.5 base model specialized in ultra-realistic human portrait photography. Trained on a highly curated dataset of 8,000 synthetic images at 640x960 resolution, it delivers photographic quality that consistently outperforms larger and more recent SD-based models in portrait-specific tasks.

Training Specs

Base architecture: Stable Diffusion 1.5 Training resolution: 640x960 Dataset size: 8,000 carefully curated synthetic images Dataset curation time: ~1 hour (high-precision intuitive selection) Specialization: Human portrait, fashion, editorial, lifestyle photography

What Makes Candy Different

Most SD 1.5 models are trained at 512x512, the native resolution of the architecture. Candy was trained at 640x960, a deliberate choice that concentrates pixel density where it matters most in portrait work — the face, hair, skin, and body proportions. The result is a model that learned human anatomy, lighting physics, and material rendering at a level of detail the base architecture was never expected to achieve.

The dataset, while small by conventional standards, was curated with surgical precision. Every image was selected to teach something specific. No filler, no noise, no redundancy. This approach — prioritizing dataset quality over volume — is the core reason Candy performs the way it does.

Photographic Quality

Candy produces results that read as real photographs across a wide range of conditions. Skin rendering includes visible pores, accurate subsurface scattering, and natural highlight behavior without the plastic smoothing common in SD 1.5 models. Hair renders with individual strand separation, correct specular response, and translucency in backlit conditions. Eyes include iris texture, accurate catchlight placement, and natural limbal ring detail.

Fabric and material rendering is physically grounded. Complex textiles including lace, satin, denim, knit, plaid, chiffon, and structured blazers all behave according to their real-world properties — draping correctly, deforming under tension, and interacting with light as expected. Fine jewelry including layered necklaces, chandelier earrings, and gemstone pendants renders with faceting, metal reflectance, and correct shadow casting on skin.

Lighting Coverage

Candy handles the full spectrum of photographic lighting conditions without degradation:

Studio soft light, harsh directional light, outdoor natural light with dappled shadows, golden hour backlight with hair translucency, sunset rooftop with cityscape, cold urban night with mixed artificial sources, warm night with string lights and lanterns, forest volumetric light with god rays, split lighting with simultaneous warm and cool sides, window light with projected shadow patterns, and interior architectural lighting.

Each lighting condition triggers appropriate automatic adaptation — skin tone shifts with color temperature, hair responds correctly to backlight, and environmental color bleeding onto skin and fabric is physically accurate.

Ethnic and Phenotypic Range

Candy was trained with genuine phenotypic diversity. It renders the following with equivalent quality across all conditions:

Northern European, Southern European and Mediterranean, Latin, East Asian, South Asian, Middle Eastern, mixed heritage, and dark skin tones including Afro-Brazilian and African phenotypes.

Skin tone adaptation is automatic and physically grounded. The model learned the real correlations between phenotype, skin undertone, and lighting response. Dark skin renders with correct specular highlight behavior, preserved shadow gradients, and accurate color without the gray shift or detail loss common in other models.

Scene and Environment Range

Beyond portrait-focused studio work, Candy performs consistently across complex environments:

Urban street scenes with architectural depth, rooftop settings with skyline, residential suburban environments, parks and tree-lined avenues with multilayer bokeh, open fields with wildflowers, dense forest with volumetric light, indoor rooms with natural window light, interior architectural spaces with columns and structured lighting, autumn outdoor scenes, and night scenes both cold and warm.

Secondary figures in backgrounds are generated without artifacts. Bokeh is organic and physically accurate, including non-circular shapes from modern LED street fixtures.

Body and Composition

Full body generation maintains correct anatomical proportions including accurate head-to-body ratio, natural waist-to-hip transition, and correct limb anatomy without distortion. Hands holding objects generate without the typical SD 1.5 finger artifacts. Midriff and exposed anatomy render cleanly without seams at clothing transitions. Multiple updo hairstyles including tight buns, low chignons, and sleek updos expose the neck and nape without distortion.

Artistic Range

Beyond photorealism, Candy demonstrates stylistic flexibility. Black and white output produces rich tonal gradients consistent with analog film photography, high contrast editorial noir, and illustrated Art Nouveau portrait styles — all while maintaining phenotypic identity and structural accuracy.

Use Cases

Fashion and editorial photography, lifestyle content, portrait generation for creative projects, dataset creation for downstream training, reference generation for artists, and virtual model photography for commercial use.

Notes

Candy is a base model. It is compatible with SD 1.5 LoRAs, ControlNet, and standard txt2img and img2img pipelines. No trigger words are required for general use. The model generalizes strongly from natural language prompts describing lighting, environment, clothing, and subject characteristics.

Trained and released by the original developer as a demonstration that dataset quality and resolution strategy matter more than parameter count.

⚠️ CONTENT WARNING (NSFW):

This merge is uncensored and capable of generating high-quality explicit NSFW content and nudity. It has a tendency towards revealing clothing in casual settings.

For SFW results: Strong negative prompts are highly recommended (e.g., nude, nipples, explicit, nsfw).

⚠️ LICENSE & PERMISSIONS (READ BEFORE DOWNLOADING)

1. PERSONAL USE ONLY This model is provided free of charge for Personal, Non-Profit, and Research use only. You may use it to create images for your personal portfolio.

2. STRICTLY NO REDISTRIBUTION

❌ DO NOT re-upload this file to Civitai, Hugging Face, or any other platform.
❌ DO NOT host this model on third-party generation services (e.g., Tensor.art, Mage.space, Telegram Bots).

3. COMMERCIAL RESTRICTIONS Using this model or its outputs for commercial revenue (Influencers, Ads, Stock Photos) without a license is PROHIBITED.

💼 COMMERCIAL SERVICES & COMMISSIONS

I do not sell the model file for commercial use. Instead, I offer premium AI solutions for brands and agencies:

✨ Exclusive AI Influencers: I create and manage consistent digital personas for Instagram/Social Media.
🏢 Corporate B2B LoRAs: Custom training for brand identity and mascots.
📸 High-End Image Packs: Monthly content packages for your brand.

To hire me for professional AI Modeling services: 📩 Contact: [[email protected]]

I recommend using the Adetailer extension.

Use this extension to fix hand errors:

https://github.com/licyk/advanced_euler_sampler_extension

Use these recommended settings for generation:

Sampling method: Euler_Max

Sampling steps: 30-50

CFG Scale: 2.0 - 7.0

Skip clip: 1-2

Sampling method: DPM++ 2M

Sampling steps: 18-30

CFG Scale: 2.0 - 7.0