Stable Diffusion Upscaling Prompt Keywords for Pixel Restoration (2026 Guide)

The SD upscaling workflow: prompts are only part of it

Important distinction before the keywords: In Stable Diffusion, prompt keywords alone don't produce 4K or 8K output files. SD generates images at a base resolution (typically 512×512 to 1024×1024) and requires a dedicated upscaling step to increase pixel count. The quality keywords signal the model to maximise detail density within that generation, but a separate upscaler is needed for true resolution increase.

The full workflow:

1. Positive prompt with quality keyword stack → tells SD to maximise detail in the generated image
2. Negative prompt with quality-degrading terms → blocks blur, noise, artifacts
3. img2img at low denoising (optional) → enhances an existing image while preserving it
4. Upscaler post-processing → increases actual pixel dimensions

This guide covers steps 1 and 2. For step 4, the recommended upscalers are: Real-ESRGAN x4+ for photography, Ultimate SD Upscale for general images, and ESRGAN 4x as a reliable fallback.

Positive prompt keyword stack for 4K/8K upscaling quality

Copy this stack and append it to any Stable Diffusion generation prompt:

For photography upscaling (photorealistic):

RAW photo, 8K ultra HD, ultra-high resolution, tack sharp, highly detailed, photorealistic, professional photography, extreme fine detail, rich color depth, cinematic clarity

For portrait pixel restoration:

RAW photo, 8K ultra HD, tack sharp, pore-level skin detail, highly detailed face, photorealistic, professional portrait photography, natural skin texture, extreme fine detail, clean sharp eyes

For landscape upscaling:

RAW photo, 8K ultra HD, ultra-high resolution, tack sharp throughout, landscape photography, highly detailed vegetation and terrain, photorealistic, cinematic dynamic range, extreme fine detail, professional outdoor photography

For product photo restoration:

8K ultra HD, commercial product photography, razor-sharp edges, highly detailed surface textures, photorealistic, accurate color reproduction, clean studio lighting, extreme fine detail, RAW photography quality

Negative prompt keyword stack (use with every upscaling prompt)

Universal upscaling negative prompt:

blurry, out of focus, soft focus, low resolution, low quality, pixelated, jpeg artifacts, noise, grain, overexposed, underexposed, poorly drawn, deformed, unrealistic, illustration, painting, cartoon, watermark, text overlay, border

Portrait restoration negative prompt (add to universal):

bad anatomy, distorted face, extra limbs, missing features, oversmoothened skin, plastic skin, airbrushed, unnatural skin texture, bad eyes, extra fingers, deformed hands

Architecture/product negative prompt (add to universal):

distorted lines, incorrect perspective, warped geometry, lens barrel distortion, chromatic aberration, vignetting

img2img settings for pixel restoration

When using img2img to enhance an existing image (rather than generating new):

For old photo restoration specifically:

• Denoising: 0.3 – 0.45 (slightly higher to allow artifact repair)
• Add to positive: restored photograph, damage repaired, reconstructed detail
• Add to negative: scratches, fading, color cast, torn, damage

Model and LoRA recommendations for pixel restoration

The prompt keywords perform differently depending on your checkpoint:

LoRA additions that boost upscaling quality:

• add_detail LoRA (weight 0.5 – 0.8): increases micro-texture rendering in face and skin
• LCM LoRA: allows fewer steps with maintained quality (faster upscaling iterations)

Complete copy-paste prompt sets for common restoration tasks

Old photo pixel restoration

Positive:

RAW photo, masterful photo restoration, 8K ultra HD, tack sharp, highly detailed, photorealistic, damage repaired, reconstructed detail, true-to-era color palette, extreme fine detail, professional archival restoration quality

Negative:

blurry, low resolution, scratches, fading, color cast, damage, artifacts, noise, grain, cartoon, illustration, unrealistic, distorted face, bad anatomy

img2img denoising: 0.4

Blurry photo to sharp 4K

Positive:

RAW photo, 4K ultra HD, tack sharp throughout, extreme fine detail, photorealistic, deblurred, crisp focus throughout, professional photography quality, rich color depth, zero motion blur

Negative:

blurry, soft focus, out of focus, low resolution, pixelated, jpeg artifacts, noise, grain, overexposed, poorly drawn

img2img denoising: 0.25 – 0.35

Portrait face restoration

Positive:

RAW photo, 8K ultra HD, tack sharp, pore-level skin detail, natural skin texture, highly detailed eyes and hair, photorealistic, professional portrait photography, true-to-life color grading, extreme fine detail

Negative:

blurry, low resolution, bad anatomy, distorted face, extra limbs, oversmoothened skin, plastic skin, airbrushed, unnatural, bad eyes, jpeg artifacts, noise, watermark

img2img denoising: 0.2 – 0.3

How these compare to ChatGPT 4K/8K prompts

Stable Diffusion and ChatGPT (DALL-E 3) are complementary tools, not direct competitors for upscaling:

The most reliable high-quality workflow combines both: use ChatGPT's 4K enhancement prompt to improve quality and guide the direction, then import the result into SD for ESRGAN upscaling to increase actual pixel count. See ChatGPT 4K Photo Enhancement Prompts for the ChatGPT side of the workflow.

Related resources

• ChatGPT 8K Ultra HD Prompts — ChatGPT-based approach to maximum quality
• Technical Prompt Keywords for High-Fidelity 8K Upscaling — Complete keyword reference across all tools
• Latest 4K Upscale AI Prompt Techniques 2026 — What changed in 2026 and what still works
• ChatGPT 4K Photo Enhancement Prompts — The photo enhancement companion for ChatGPT
• AI Upscale Image to 4K: Sample Prompts — Side-by-side comparison of generative vs dedicated upscaling