SoonLab 2.0 is Here — Create Games with Your AI Agent, Extra 10% OFF · Code: SLTOP10
Help Guide
Everything you need to know about building, publishing, and remixing games on SoonLab.
On this page
What is SoonLab?
SoonLab is an AI-powered game creation engine that lets you build fully playable games using natural language. No coding or complex tools required — simply describe the game you want to create, and SoonLab's AI builds it in real time.
The platform is for everyone — from first-time creators with no technical background to large professional studios looking to prototype games quickly. You own everything you create, and you're free to publish, share, and monetize your games as you see fit.
How It Works
The core workflow is conversational. You describe what you want, the AI builds it, you review the results, and then iterate through follow-up instructions until the game matches your vision. Think of it as a creative collaboration with an AI game developer.
You can create everything from simple arcade games to complex interactive worlds — it all depends on your prompts and how you iterate. There's no limit to the types of games you can build: the more detail you provide, the better the results.
Creating Games
Starting a New Game
On the home page or the create game page, enter your game idea into the prompt box and click Create / Generate. This is the primary entry point for creating games in SoonLab. Each new game creation consumes credits from your account.
How Prompts Work
A prompt is how you communicate your idea to the AI. The more specific and detailed your prompt, the better the result. Think of a prompt as a creative brief — it should clearly convey the game genre, art style, gameplay mechanics, and what makes it unique.
Good prompts typically include:
- Genre and perspective — e.g., "top-down RPG," "side-scrolling platformer," "first-person puzzle game"
- Art style — e.g., "pixel art," "hand-drawn watercolor," "dark sci-fi," "vibrant cartoon"
- Core mechanics — e.g., "collect coins and avoid enemies," "solve physics-based puzzles," "build and manage a farm"
- Setting and story — e.g., "a haunted medieval castle," "an underwater city from the year 1000 AD"
- Win/lose conditions — e.g., "survive 5 waves of enemies," "reach the exit before time runs out"
Reference Materials and Style Guides
You can upload reference images alongside your prompt to give the AI visual guidance. This is especially useful for defining art styles, color palettes, or character designs for your game.
Supported reference types:
- Style references — Upload images that represent the overall visual style you want (concept art, screenshots from other games, mood boards).
- Character references — Upload character images you'd like the AI to use as inspiration for game characters.
- General references — Any visual material that helps convey your design vision, such as UI mockups or environment sketches.
Use the image upload button in the chat input area. You can upload multiple images at once.
Iterative Building
Creating a game on SoonLab is an iterative process. No game comes out perfect on the first try. After the initial version is generated, you can continue sending messages to refine the game, add features, fix issues, or even change direction entirely. Each modification consumes credits from your account.
You can use follow-up prompts to:
- Add new levels, characters, or mechanics
- Change the art style or color palette
- Fix bugs or adjust difficulty
- Add sound effects or background music
- Improve controls and player movement
- Add menus, UI elements, or score tracking
Debugging and Issue Fixes
Bugs are a normal part of game development — especially when using AI that builds games from natural language. SoonLab makes it easy to find, describe, and fix bugs through the same conversational workflow you use to create games.
Finding Bugs During Playtesting
You can catch most issues in the real-time preview within the Studio editor. Play your game like a player would — try every interaction, push the boundaries of the rules, and deliberately do things that "shouldn't" work. The more thoroughly you test, the more bugs you'll catch before your players do.
What to watch for:
- Visual glitches — sprite overlap, missing assets, broken layouts, or incorrect colors
- Mechanic failures — actions not working, poor control feel, or rules triggering at the wrong time
- Game flow issues — getting stuck on a screen, win/lose conditions not triggering, or progress not saving
- Performance problems — stuttering, lag, or screen freezing during gameplay
- Edge cases — scoring zero points, holding two keys simultaneously, or abnormal behavior when hitting a wall mid-jump
How to Describe a Bug Effectively
The best way to help the AI fix a bug is to describe it clearly. A vague "it's broken" leaves the AI guessing; a precise description lets it zero in on the root cause.
Good bug reports typically include:
- What you were doing — the exact action or sequence that triggered the bug (e.g., "I pressed the jump key while standing next to a wall")
- What you expected to happen — the intended behavior
- What actually happened — the incorrect behavior, described in as much detail as possible
- Where it happened — which screen, level, menu, or UI element
- How often — every time, sometimes, or only after performing a specific action
Asking the AI to Fix Bugs
Once you've described the bug, simply ask the AI to fix it in the chat. No special commands needed — just send the bug report like a regular message. The AI will analyze your game, locate the issue, and fix it. You can then immediately re-test in the preview.
For trickier bugs, you can ask the AI to investigate first — for example, "Before changing anything, investigate why X is happening." The AI will explain what it thinks is wrong before making any modifications, so you can confirm or adjust the approach.
When the First Fix Doesn't Work
Sometimes a fix doesn't fully solve the problem or even introduces new ones. That's normal. Just reply with what you observed after testing — for example, "The character isn't stuck anymore, but now they fall through the platform." Each follow-up gives the AI more information to narrow down the issue and find the real cause.

