Brainiall Pronunciation
Phone-level scores that exceed human agreement
The only commercial pronunciation API where the model is more consistent than trained human raters. Phone PCC 0.590 on our production Light tier and 0.682 on the Premium fine-tuned tier — both above the 0.555 inter-annotator ceiling on our public test set. AWS and GCP have no equivalent product; Azure offers 33 locales without publishing accuracy numbers.
Interactive Phoneme Scoring Demo
Click on individual phonemes below to inspect acoustic scores, ARPAbet alignment, and diagnostic feedback generated at runtime.
Why this scoring matters
Traditional speech apps rely on vague sentence-level grades or subjective human reviews. Human raters only achieve a 0.555 Pearson Correlation Coefficient (PCC) with each other. Brainiall delivers objective, phone-level scoring that exceeds human agreement ceilings.
Two-line integration
Send raw audio (base64) plus the expected text. Get back per-phone, per-word and per-sentence scores — one HTTP call, no streaming required.
curl -X POST https://api.brainiall.com/v1/pronunciation/assess \
-H "Authorization: Bearer $BRAINIALL_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"audio": "<base64-wav-or-mp3>",
"text": "She sells seashells by the seashore",
"tier": "premium",
"locale": "en-US"
}'
# Response (deterministic score breakdown):
# {
# "sentenceScore": 86,
# "accuracyScore": 88,
# "fluencyScore": 84,
# "completenessScore": 100,
# "wordScores": [
# { "word": "she", "score": 96, "phonemes": [{ "phone": "SH", "score": 98 }, { "phone": "IY", "score": 94 }] },
# { "word": "sells", "score": 84, "phonemes": [{ "phone": "S", "score": 92 }, { "phone": "EH", "score": 88 }, { "phone": "L", "score": 81 }, { "phone": "Z", "score": 75 }] }
# ],
# "latency_ms": 382
# }When NOT to Use Pronunciation Assessment
- Open-ended unstructured conversations without reference text: Pronunciation scoring requires the target text to compare acoustic phonemes against canonical alignment. For freeform transcription, use our Streaming STT engine in the Audio family.
- Clinical motor speech disorder diagnosis: While acoustic scores provide objective metrics, clinical neurological evaluations require a licensed speech-language pathologist.
- Tonal pitch or musical singing grading: Our scoring models are optimized for natural spoken linguistic phonemes, not musical melody or operatic pitch scales.
Methodology
The Light tier is a 9-model ensemble (4× MLP + 4× XGBoost + 1× PhoneTransformer) over features extracted from a Brainiall Speech (Edge tier) acoustic model — same backbone we use for our STT product. The Premium tier adds a fine-tuned Brainiall Speech (Premium) engine with a learnable weighted-sum across its 24 transformer layers, plus an ordinal-contrastive scoring objective. All numbers come from the public benchmark methodology page; the test set is our public test set (the SOTA reference for L2-English pronunciation).
Pricing
Light tier scores at production-grade quality with CPU inference; Premium tier uses a fine-tuned Brainiall Speech (Premium) engine for the highest published accuracy on our public test set. Discount rule: 50% off market average when our quality is superior — which it is, on both tiers.
Free
100 minutes/month · Light tier · Forever free
Starter
1,000 minutes/month · Light tier · Word + sentence scores
Pro
10,000 minutes/month · Light + Premium · Phone-level scores · 99.5% SLA
Business
50,000 minutes/month · Dedicated capacity · Direct line to founder