SpeakDuo AI models demo

Record 4 to 15 seconds of English (or use the sample clip). The clip goes to a fine-tuned Gemma 4 model on a GPU; answers take about 2 to 8 seconds.

Read this aloud

Model

Sentence you read (models 2 and 3 score it against this text; edit if you said something else)

Target accent (model 2)

Samples averaged (more = steadier scores, slower)