A calmer practice loop
Choose. Practise. Understand. Repeat.
No complicated setup. Each practice begins with a small, approachable target and ends with a clear next move you can read rather than hear.
Pick your practice
Choose a phoneme, word or sentence from your learning path. New learners start with sounds that are visible on the face and common in everyday words, so you can check your own work in a mirror.
See the cue before you try
Every target opens with an articulation cue: where the tongue sits, whether the lips round, whether air is stopped or flowing, and whether the throat buzzes. You know what you are aiming for before you record anything.
Record when ready
No timer, no countdown. The microphone is captured raw — noise suppression, echo cancellation and auto-gain are switched off — because those “helpful” filters erase exactly the details a deaf learner needs measured.
Read your feedback
A phoneme model with no language model reports what you actually said — g ə instead of k ə — never a corrected guess. You get a score per sound, the physical reason, and one thing to change next. Nothing is audio-only.
Follow the pattern
Your attempt history collects quietly. Over a week or two it shows which sounds are reliable and which deserve another look, so practice time goes where it matters.
Build the habit
Short and frequent beats long and occasional. Eight focused minutes a day outperform one long weekend session, and a session can end cleanly at any point without losing your place.
Save a reference pronunciation
On any practice screen, a therapist or family member can record the target. Future attempts also get a similarity score against that voice — robust to atypical voice quality.
Match, confidence, similarity
Phoneme alignment tells you what was substituted. A per-sound confidence grades it 0–100. Embedding similarity compares you to the reference. Together they are hard to fool and easy to read.
The analysis in depthMeasurements you can feel
Voice onset time, voiced fraction, pitch, aspiration noise ratio and vowel formants — turned into “hand on throat” and “palm at lips” instructions instead of graphs.
Try one loop
It takes about two minutes.
Open Sign & Talk in your browser. The Android app follows the same account, so nothing is lost when you switch.