I Built a Diwali Game Using AI in Just 1 Day – MSI Prestige, RTX 5070 & RTX 5090

Can you build a Diwali game using AI without a full game development team? That was the challenge behind one of my most interesting Tech Dreams experiments: start with a playable 2D concept, create custom festive artwork and finish with a game that actually feels like Diwali—all within a one-day video challenge.

Instead of relying on one powerful computer for every task, I used three different MSI machines. The idea was to demonstrate how AI-assisted coding, local AI image generation and high-end GPU workflows fit into different stages of making a game.

Watch the Tech Dreams video: [Embed the original YouTube video here.]

The Game Concept: Collect Diyas, Avoid Obstacles, Celebrate

The plan was a straightforward 2D game set in a colourful Indian festive environment. The player moves around the scene, collects scattered diyas, avoids obstacles and tries to complete the objective before the timer runs out. Each collected diya contributes to the score. Completing the challenge triggers celebratory lights, fireworks and an ending message.

The scope was deliberately small: no complex 3D engine, no enormous open world and no attempt to replace a professional development team. The interesting question was whether AI tools could make a tightly scoped creative project faster to build and easier to iterate.

Stage 1: Claude Code and MSI Prestige 13 AI+ EVO

I began on the MSI Prestige 13 AI+ EVO, a lightweight productivity laptop using an Intel Core Ultra 9 288V processor, 32GB LPDDR5X memory and a 1TB NVMe SSD in the configuration shown in the video.

For the development workflow, I used Claude Code with Python and Pygame. Rather than writing every routine manually, I described what the game needed: keyboard movement, collisions, collectable objects, scoring, a timer and a playable loop. Claude Code helped create and modify project files while I ran the game, checked the output and fixed problems through repeated testing.

That last step matters. An AI coding assistant can produce a useful first draft, but the developer still needs to verify behaviour, reproduce errors and decide whether the result matches the intended design.

At the end of the first stage, the game worked. It was not particularly attractive: temporary shapes and a simple background stood in for finished artwork. However, the core gameplay loop was in place.

The takeaway: coding a modest 2D prototype does not automatically require a top-end discrete GPU. A responsive processor, adequate memory and a usable development environment may be more relevant to this stage.

Stage 2: Creating Game Assets with ComfyUI and RTX 5070

The next challenge was visual identity. For this part of the video, the featured machine was the MSI Crosshair 16 HX AI, using an NVIDIA GeForce RTX 5070 Laptop GPU.

I used ComfyUI, a node-based interface for generative AI workflows, to create suitable festive artwork. The target look included an Indian street with colourful houses, decorative lights, rangoli, diyas, gift boxes, characters and other environment elements.

Local image generation can benefit from a dedicated GPU because model inference is computationally demanding and video memory affects which models, resolutions and workflows are practical. The RTX 5070 Laptop GPU configuration featured in the video has 8GB GDDR7 VRAM, making memory-conscious settings an important part of the process.

The generated images were not the finished game by themselves. Assets had to be selected, adapted and integrated into the prototype. I then returned to the coding workflow to replace placeholders and make the visual components work with the existing game logic.

Once the new graphics were in place, the game felt substantially closer to the Diwali concept that I originally imagined.

Stage 3: Refining the Experience on an RTX 5090 Desktop

For the final stage, I switched to an MSI MEG desktop build featuring an AMD Ryzen 9 9900X3D and a GeForce RTX 5090 Lightning Z. The desktop setup shown in the video also included an MSI motherboard, high-end cooling, fast SSD storage and an ultrawide QD-OLED monitor.

The RTX 5090’s large graphics-memory capacity makes it interesting for demanding local creative workflows. Here, the emphasis moved to refining selected visual assets, working with higher-resolution material, polishing the presentation and preparing the game’s festive finale.

I also explored LM Studio for running a language model locally and generating short celebratory messages for the end screen. This was a different use of AI from the ComfyUI workflow: language generation rather than graphics generation.

The final playable game combined the code from stage one, artwork from stage two and additional polish from the desktop stage. Fireworks, festive visual effects and a clear completion sequence made the simple collection mechanic feel more rewarding.

Do You Need an RTX 5090 to Make a 2D Game?

Absolutely not. That is one of the central lessons of this experiment.

A basic Python game can be developed on a capable everyday computer. A powerful GPU becomes more relevant when you intentionally add local generative AI, complex image processing or heavier creative workloads. The correct hardware depends on the task—not on the assumption that every stage of development needs the most expensive machine.

What This AI Game Development Experiment Taught Me

The biggest advantage of the three-device approach was seeing the project broken into manageable jobs:

• MSI Prestige 13 AI+ EVO: prototype the rules, movement and scoring with Claude Code.
• MSI Crosshair 16 HX AI: develop custom Diwali visuals using local AI image generation.
• MSI RTX 5090 desktop: explore heavier creative processing and refine the final presentation.

The experiment also demonstrated the limitations of AI-assisted development. Prompting alone is not enough. You still need to test the game, evaluate assets, catch bugs and make creative decisions.

Final Verdict

This was a practical example of using AI tools to build a Diwali game from scratch, rather than simply discussing what AI might be able to do. The finished project was a modest 2D game, but it offered a useful demonstration of coding assistance, visual asset generation and local AI inference working together.

Watch the original Tech Dreams video for the gameplay demonstration and the full three-machine workflow. If you are experimenting with AI game development, start with one small, playable idea and add complexity only after the basics work.