What prompt makes a basic AI game, and where do you run it?
For a first game, describe one action and one obstacle: “a forest adventure where you gather glow seeds and avoid shadow beasts.” Paste it into the WordstoGame AI game generator, choose Collector and Fantasy, and press Generate game. The game runs on the same page in your browser. You do not need to paste code into another app.
Try your first promptThe real question behind this walkthrough
In a discussion about making games with AI, a beginner asked what instruction produces a basic game and where to run the result. This page answers those two practical steps for WordstoGame; it does not rank AI models.
“Then with the command recieved from the AI, where to put it for it to work.”
The exact workflow we tested
We used the live homepage, rather than a mockup, and entered the 70-character forest prompt above. Here are the settings and the sequence needed to reproduce the setup:
- Open the homepage tool and enter the sentence in Describe your game. The field accepts one idea of 3–240 characters.
- Set Game type to Collector. Set Visual style to Fantasy. We left Difficulty on Normal and all Extras off.
- Press Generate game. Our first visible result was titled Forest Adventure Gather and displayed LOCAL ENGINE. Its objective was to collect 16 glow seeds across 5 waves.
- The optional model refinement then changed our title to Glowseed Quest. The result badge became MODEL + LOCAL ENGINE. The type, style, difficulty, wave count and 16-pickup objective remained the same in this run.
- Use the displayed movement controls: arrows or WASD on a keyboard, or drag on the game on touch devices. Read the objective and progress counter before trying to avoid the moving hazards.
- Press Copy brief to record the result, then Copy link to keep an address that can reopen its settings. We tested both buttons and reopened the copied address.
What our generated result actually contained
- Title after refinement
- Glowseed Quest
- Settings
- Collector · Fantasy · Normal · No extras
- Objective
- Clear 5 waves: collect 16 Glow Seeds in total.
- First-wave counter
- 0/2 pickups at the start; the 16-pickup total spans the waves.
- Observed run result
- We left the initial run idle while recording the output. It ended with “You lost,” on wave 1 with score 0. This test confirms a running game and a loss state; it is not a completed five-wave playthrough.
The copied brief included the win condition, “all 5 waves are cleared,” and the lose condition, “health reaches zero.” It also described drifting, chasing and orbiting hazards, with a three-hit elite from wave 3. We did not reach wave 3 in this test, so that last detail is a brief description, not an observed enemy encounter.
Where to run and reopen this example
Open the tested forest collector settings. The address opens the homepage with Collector, Fantasy, Normal and the forest theme selected. In our reopen test, the page explicitly reported that it was built locally with no request sent.
Reopening changed the title back to Forest Adventure Gather, with LOCAL ENGINE displayed. The objective still required 16 glow seeds in 5 waves, and progress started at 0/2 with score 0. This is a fresh run, not a saved session. The link carries settings; it does not preserve the model's exact title, colours, your original custom sentence or your progress. The forest sentence appeared again here because it is the built-in prompt for that theme.
How to adapt the prompt without asking for too much
Keep the first request within one of the three supported types: Shooter, Collector or Survival dodger. Specify the action, the setting and what to avoid. Use the dropdowns if automatic interpretation chooses a different type than you expected. Each game still uses simple 2D arcade rules.
A fantasy theme is a visual choice, not a full RPG system. WordstoGame does not currently create 3D worlds, multiplayer sessions, branching text adventures or downloadable game files. Copy brief is text; Copy link is a browser address. For privacy details about optional model refinement and the local fallback, read the privacy page.
Start with the tested collector, then change one setting and regenerate. That makes it easier to see what your input changed and whether the result fits the small game you intended.