Executing a single action (e.g., a manual button press) while concurrently inhibiting another, prepotent action (e.g., a saccade) typically results in frequent false-positive executions of the latter action. Executing both actions together does not require inhibitory control and can thus be less error-prone, creating a relative dual-action performance benefit. We here examined the effect of stimulus-response (S-R) translation automaticity on the frequency of false-positive errors in single-action trials and accuracy-based dual-action benefits. Based on a color cue, participants responded with either a single saccade, a single manual button press, or both actions together to a single visual target stimulus. Stimulus type was varied between blocks with decreasing translation automaticity into a spatial (left vs. right) response (peripheral square > central arrow > arbitrary central shape). This order was reflected by a corresponding increase in correct RTs. False-positive errors in single-action trials indicative of inhibition failures were less frequent in the central arrow condition compared to the peripheral square condition. Inhibition failures were most frequent in the arbitrary central shape condition, however, so were overall error rates. The relative dual-action benefit thus decreased as a function of S-R translation automaticity. These results suggest better inhibitory control of a prepotent action when a concurrently executed action takes longer (vs. shorter) to be specified. However, this only holds when specification is not overly difficult.