Blend word (SuperSpeech test)

Isolated endpoint: POST /api/testing/speech/blend/word (multipart: audio, word). The server converts M4A / WebM / MP3 / … to 16 kHz mono WAV (FFmpeg), labels the file as audio/wav for the API, then calls SuperSpeech. For a single word, say it as one smooth blend — not separate beats for each letter. Use the preset dropdown or type any word (2–64 letters; API keeps letters only, e.g. hello → /HELLO/).

Preset words (optional) Target word

Quick picks — full list in dropdown

Dictionary

Age group

Slack (higher = more timing tolerance, try 0.65–0.85)

Slow down WAV (FFmpeg atempo)

Gain (dB, optional — if you still see “volume too low”)

Or upload a file

Response

No result yet.