Speech AI MCP Server

AI-powered speech tools for MCP-enabled AI agents: pronunciation scoring, speech-to-text, text-to-speech, and multilingual transcription.

Tools

Tool	Description
assess_pronunciation	Score English pronunciation from audio (0-100 at overall, sentence, word, phoneme levels)
transcribe_audio	Convert spoken English to text with word-level timestamps
synthesize_speech	Generate natural speech from text (12 voices, American & British)
transcribe_audio_pro	Whisper Large V3 Turbo: 99 languages, speaker diarization
list_tts_voices	List available text-to-speech voices
check_pronunciation_service	Health check for pronunciation backend
check_stt_service	Health check for STT backend
check_tts_service	Health check for TTS backend
check_whisper_service	Health check for Whisper backend

Pronunciation Scoring

Returns scores (0-100) at four granularity levels:

Level	Description
Overall	Global pronunciation quality
Sentence	Sentence-level fluency and accuracy
Word	Per-word pronunciation scores
Phoneme	Individual sound accuracy (IPA + ARPAbet)

Performance

• Accuracy: Exceeds human inter-annotator agreement (PCC 0.576 vs 0.555)
• Validated: 9,259 utterances across 7 L1 backgrounds, zero errors
• Latency: Sub-300ms for pronunciation and STT

How to Use

MCP Endpoint

https://Ym2gS88TksnTdTcPq.apify.actor/mcp?token=YOUR_APIFY_TOKEN

Example: Pronunciation Assessment

{
  "audio_base64": "<base64-encoded-audio>",
  "text": "The quick brown fox jumps over the lazy dog"
}

Example: Text-to-Speech

{
  "text": "Hello, how are you today?",
  "voice": "af_heart",
  "speed": 1.0
}

Pricing

$0.02 per tool call (pay-per-event).

Technical Details

• Pronunciation Model: Conformer-CTC Small (17MB, INT8 quantized)
• TTS Model: Kokoro-82M (12 English voices, 24kHz WAV)
• STT Pro: Whisper Large V3 Turbo (99 languages, speaker diarization)
• Audio: Supports WAV, MP3, OGG, FLAC, WebM
• Backend: Azure Container Apps, auto-scaling

Links

• Demo: https://huggingface.co/spaces/fabiosuizu/pronunciation-assessment
• API Docs: https://apim-ai-apis.azure-api.net/pronunciation/docs
• Company: https://brainiall.com