Could computer vision play Automaton Attack reliably?

November 20, 2026 3 answers
Asked by deckCurve+Rhubarb

I'm curious about using computer vision to read the targets and automate typing in Automaton Attack. How would that compare with a script that repeatedly sends every letter, and what would make detection unreliable?

Answers

DOSsoundfont_2B

A vision approach could capture the game screen, locate the words, read them, and send the detected letters as inputs. The tricky part is that overlapping words can confuse text recognition, so a workable system may still miss or misread targets.

For a higher score, it would be worth measuring where errors happen. Separate missed text from incorrect recognition and input timing issues before changing the whole pipeline.

quickcast_3L!:

Overlaps seem especially awkward because a reader might interpret parts of two words as one. Tracking detections across frames could be something to explore.

Reply to this answer
rankedQueue+Slushie

If the goal is just to automate inputs, repeatedly sending all the letters is a simpler approach. Vision is a different challenge: it tries to react to what's actually on screen, but adds recognition errors and more moving parts.

Reply to this answer
kartSlipstream+Lunchpail

I'd treat score and detection quality as separate measurements. A system might recognize most words correctly but still score poorly if it responds too slowly. Compare runs under similar conditions and check whether overlapping text accounts for the misses.

Reply to this answer

Add your answer

Answers are reviewed before publication.

Related Discussions