Since both children are still only learning to read and write, I dispensed of all the text, and use only pictograms and speech synthesis for communication with the user. Both games run in a browser. I tested only with Chrome, but any browser which supports speech API should work.

The first game trains the passive vocabulary. The user listens to the word and is presented two pictograms to choose from, a correct one and an incorrect one. For selection I use body pose estimation. Raising the left hand chooses the left pictogram and, analogously, raising the right hand selects the right pictogram. So far, it seems that kids prefer body movements to typing / clicking on the screen, especially the younger one.

The second game is for practicing the active vocabulary. The user is presented with a pictogram and needs to pronounce the word. Three attempts are allowed, and if the user does not know the word, it needs to be later repeated. This is basically a flashcard method coupled with speech recognition.

In both games, the user gains diamonds. I inserted a silly ghost which tries to steal them to add some sense of temporal urgency.

You can access the games at: https://github.com/mristin/dance-to-english/

and

https://github.com/mristin/say-in-english/

Any feedback, suggestions or invitations for collaboration are very welcome!