Click or drag & drop audio here WAV, MP3, FLAC, OGG — any format
▶ Replay · detected paralinguistic events shown in place
0:000:00
🔄 Gender
📅 Age Group
Correct this result▾
The model is weakest on real voice notes in under-represented languages —
corrections are stored with the audio and become training data. Only change what's wrong.
Detected events — tap any that are wrong
Missed an event? Add it
no changes yet
🎙 Mic comparison — same sentence, two microphones
The same voice can read differently depending on the microphone —
that is a real failure mode we are tracking. Say the same sentence into each mic.
Both takes and their predictions are saved as one linked pair.
1 First microphone
00:00
no take yet
2 Second microphone
00:00
no take yet
😃 Emotion
🔊 Paralinguistics — Detected Events (0)
Frame-level detection & tagging — every sound with its onset → offset. This is the model's primary, most reliable paralinguistic output.
Sound
Onset
Offset
Confidence
When
No sound events detected
🏷️ Whole-clip classification — one tag for the entire clip (weaker; out-of-domain on long voice notes)