How AI interpreted the description: two test rounds
Actual Speakasso test results from 22 September 2026: two paintings with two descriptions each. These are prerecorded test scenarios, not user matches. Images came from the real AI pipeline on an isolated test setup. Scores belong to that game version and are not objective similarity percentages.
Which description wins?
Look at the painting and choose the description you think earned more points.
This replays a Speakasso test from 22 September 2026. Your choice reveals saved results; no new image is generated.
Visual score: 73. Penalty: 0. Final score: 73.
Visual score: 60. Penalty: 45. Final score: 15.
The second image scored 60, but forbidden words cost another 45 points. Both the image and the rules affect the final score.
These are the saved test scores, not a score for your own play.
The bedroom: similarity and word penalties
The first description earned 73 points. The second received an image score of 60 and a 45-point penalty, finishing at 15. This completed test shows how word penalties can change the outcome.
Read both descriptions and see the results
The rainy street: an ambiguous word
One image contains an architectural dome instead of an umbrella. The other has an umbrella, but the people face away. The original round was voided and the images were judged separately. The full case explains these limits.
See the exact wording and both results
Try it in Speakasso
Try the painting description exercise, then compare your words with an example. Speakasso is on Google Play.