Key Points
- Gemini 3 enables conversational “vibe coding” to create HTML game demos.
- Users produced demakes of Chip’s Challenge, Jezzball, 8‑bit Tomba!, and Silent Hill.
- The “Thinking” model yields deeper code generation than the faster variant.
- Gemini supplies step‑by‑step guidance for packaging games into other formats.
- Compared to free ChatGPT, Gemini delivered more polished visual and audio elements.
- AI‑generated demakes offer a low‑cost way to revisit classic games.
Gemini 3 Introduces Vibe Coding for Game Creation
Google’s latest large language model, Gemini 3, offers a conversational workflow that lets users ask the AI to generate simple video games. The process, described as “vibe coding,” involves a back‑and‑forth dialogue where the model refines code based on user feedback, ultimately producing HTML files that run directly in a web browser.
From Concept to Playable Demos
Using this method, several classic titles were recreated as stripped‑down demakes. The AI produced a Chip’s Challenge‑style maze runner, a functional Jezzball ball‑bouncing puzzle, an 8‑bit rendition of the PlayStation classic Tomba!, and a minimalist Silent Hill experience featuring fog, ambient sounds, and basic puzzles. Each demo captured the core mechanics of its inspiration while remaining lightweight enough for immediate play.
Guidance on Model Selection and Output
When configuring Gemini, users are advised to select the “Thinking” model rather than the faster “Fast” option. The Thinking mode provides more detailed code and explanations, which is crucial for generating complete game files. The AI also supplies instructions for converting the HTML output into other formats, such as Android APK packages, though it cannot perform the conversion itself.
Comparison With Other AI Tools
In a side‑by‑side test, the free version of ChatGPT was asked to create an 8‑bit Tomba! demo. While it produced a playable result, the styling and polish fell short of Gemini’s output. ChatGPT required additional prompting to adjust visuals, whereas Gemini incorporated aesthetic details more seamlessly across iterations.
Implications for Retro Gaming Preservation
The ability to generate quick, playable versions of older games could help enthusiasts experience nostalgic titles without the expense of original hardware or rare software copies. By leveraging the vast text and code data that Gemini has been trained on, developers and hobbyists can recreate the essence of vintage games in a modern, accessible format.
Source: cnet.com