Remaking Old Computer Graphics With AI Image Generation

2023-01-01 · Source: Jay Alammar · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Gaming & Interactive Media · Depth: Intermediate, long

Summary

This analysis explores using AI image generation tools like Stable Diffusion, DALL-E, and Midjourney to remake graphics from the 1987 video game "Nemesis 2 on the MSX" into higher-fidelity versions. The author details the process of generating eight cinematic panels, highlighting the iterative nature of prompt engineering and the varying capabilities of each tool. Stable Diffusion, accessed via Dream Studio, required extensive prompt tweaking and style keywords, often sourced from galleries like Lexica. Midjourney consistently produced high-quality, aesthetically pleasing results with simpler prompts, though sometimes sacrificing fidelity to the original image's essence. DALL-E demonstrated strong inpainting and outpainting capabilities for expanding images and manipulating specific elements. The post also provides a comparative review of the pros and cons of Dream Studio, Midjourney, and DALL-E based on user experience, API access, UI, and image quality.

Key takeaway

For AI Engineers or graphic designers looking to modernize legacy visual assets, you should experiment with multiple AI image generation platforms. Start with Midjourney for high aesthetic quality, but be prepared to use Stable Diffusion (via Dream Studio) for more control over specific styles and DALL-E for precise image manipulation like inpainting or outpainting. Expect an iterative process of prompt refinement and consider external tools like Photoshop for elements like text or exact placement.

Key insights

AI image generation can remake old graphics, but requires iterative prompting and tool-specific techniques.

Principles

• Prompt engineering is crucial for style control.
• Different AI models excel in distinct areas.
• Iterative refinement is key for desired output.

Method

The workflow involves initial subject-based prompting, refining with style keywords from galleries like Lexica, and using inpainting/outpainting for specific element manipulation or canvas expansion.

In practice

• Use Lexica for prompt inspiration.
• Try Midjourney for high-quality aesthetics.
• Employ DALL-E for inpainting/outpainting.

Topics

• Generative AI
• Prompt Engineering
• Stable Diffusion
• Midjourney
• DALL-E

Code references

• openai/glide-text2im

Best for: AI Engineer, Machine Learning Engineer, Prompt Engineer

Related on AIssential

• The Complete Guide to AI Art Styles (100+ Prompt Ideas) — Effective AI image generation relies on combining visual medium, style, lighting, and ratio for diverse, tool-agnostic artistic outputs.
• Ideogram and Reve rethink how AI images get made — AI image generation is evolving from text prompts to post-creation layout and granular editing for greater user control.
• 5 steps to generate consistent brand images with Midjourney — Consistent AI image generation relies on structured workflows, visual language, and iterative refinement.
• Claude’s Hidden Art Skill: Making Illustrations With Code — Claude generates editable, scalable vector graphics (SVG) from text prompts, offering unique iterative refinement capabilities.
• A picture is worth a thousand words for an LLM — Mood boards and style references are crucial for guiding AI image generation towards a specific aesthetic.
• EPEdit: Redefining Image Editing with Generative AI and User-Centric Design — EPEdit offers cost-effective, user-friendly generative AI image editing via zero-shot Stable Diffusion, eliminating fine-tuning.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Jay Alammar.