メインコンテンツへスキップ
Brainiall
サインイン始める
🎙️ 音声・音響ファミリー

音声・音響 インテリジェンスAPI

低遅延WebSocketストリーミング音声認識、高精度話者分離、声紋バイオメトリクス、スタジオ品質の自動吹き替え。

無料トライアルを開始

リアルタイムストリーミング音声認識

WSS / POST/v1/audio/stream & /v1/audio/transcribe

250ミリ秒未満のチャンク遅延と単語単位のタイムスタンプを備えた双方向WebSocket音声認識。

  • Sub-250ms chunk latency with WebSocket streaming protocol
  • Word-level timestamps with punctuate-and-capitalize normalization
  • Multilingual acoustic coverage across 99 spoken languages
ProtocolWebSocket / HTTP
Latency P95< 250ms
Audio InputsPCM, WAV, Opus, MP3
WebSocket Audio StreamNode.js / Browser
const ws = new WebSocket("wss://api.brainiall.com/v1/audio/stream", {
  headers: { Authorization: `Bearer ${API_KEY}` }
});

ws.send(JSON.stringify({ config: { language: "en", interim_results: true } }));
audioSource.on("data", (chunk) => ws.send(chunk));

ws.on("message", (data) => {
  const result = JSON.parse(data);
  console.log("Transcript:", result.text, "Confidence:", result.confidence);
});

話者分離（ダイアライゼーション）

POST/v1/audio/diarize

複数人の会議や通話をサブ秒単位の境界精度で個別の発話ターンに自動分割します。

  • Accurate “who spoke when” segmentation with sub-second boundaries
  • Handles overlapping speech, interruptions, and variable speaker counts
  • Emits structured turns with speaker labels (SPEAKER_00, SPEAKER_01) and time ranges
Speaker LimitUp to 24 speakers
Accuracy DER< 9.2% on AMI test
POST /v1/audio/diarizecURL
curl -X POST https://api.brainiall.com/v1/audio/diarize \
  -H "Authorization: Bearer $BRAINIALL_KEY" \
  -F "audio=@meeting_record.wav" \
  -F "num_speakers=2"

# Response:
# {
#   "segments": [
#     { "speaker": "SPEAKER_00", "start": 0.0, "end": 4.12, "text": "Good morning team." },
#     { "speaker": "SPEAKER_01", "start": 4.25, "end": 8.70, "text": "Morning, let us review Q3." }
#   ]
# }

声紋照合・話者特定

POST/v1/audio/voice-id/verify & identify

周囲のノイズに強い1:1の生体音声認証および1:Nの話者特定ベクター照合。

  • Biometric 1:1 voice verification against enrolled voiceprints
  • 1:N speaker identification over millions of customer profile vectors
  • Text-independent acoustic embeddings resilient to background acoustic noise
Enrollment Time3 to 5 seconds audio
EER< 0.9% on VoxCeleb1
POST /v1/audio/voice-id/verifycURL
curl -X POST https://api.brainiall.com/v1/audio/voice-id/verify \
  -H "Authorization: Bearer $BRAINIALL_KEY" \
  -F "user_id=usr_8921" \
  -F "audio=@auth_sample.wav"

# Response:
# { "match": true, "similarity": 0.942, "threshold": 0.85, "verified": true }

音響補正・ノイズ除去

POST/v1/audio/enhance

空調音や反響の除去、スタジオ品質の音声分離、48 kHzへの音質復元。

  • Suppresses HVAC, keyboard clicks, traffic rumble, and background chatter
  • Removes room reverberation for clean, up-front vocal presence
  • Outputs broadcast-ready 48 kHz normalized audio
POST /v1/audio/enhancecURL
curl -X POST https://api.brainiall.com/v1/audio/enhance \
  -H "Authorization: Bearer $BRAINIALL_KEY" \
  -F "audio=@noisy_mic.wav" \
  -F "dereverb=true" \
  --output enhanced_studio.wav

自動吹き替え

POST/v1/audio/dub

元の声質と感情を維持しながらリップシンクに合わせた多言語動画・ポッドキャスト吹き替え。

  • End-to-end automated video and podcast dubbing into 40+ languages
  • Preserves original vocal tone, emotion, and speaker identity across languages
  • Automatic timestamp duration alignment with video lip-sync pacing
POST /v1/audio/dubcURL
curl -X POST https://api.brainiall.com/v1/audio/dub \
  -H "Authorization: Bearer $BRAINIALL_KEY" \
  -F "video=@course_video.mp4" \
  -F "target_language=es" \
  -F "voice_match=true" \
  --output dubbed_spanish.mp4

音声直接翻訳（Speech-to-Speech）

POST/v1/audio/speech-to-speech

テキストの中間変換を挟まず、話された音声を対象言語の音声へ直接変換します。

  • Spoken audio in, translated spoken audio out in a single orchestrated pass
  • Streaming audio translation without intermediate text roundtrips
POST /v1/audio/speech-to-speechcURL
curl -X POST https://api.brainiall.com/v1/audio/speech-to-speech \
  -H "Authorization: Bearer $BRAINIALL_KEY" \
  -F "audio=@english_query.wav" \
  -F "source_lang=en" \
  -F "target_lang=pt-br" \
  --output translated_speech.wav

統合Speech Suite Proパイプライン

POST/v1/audio/pipeline

音声認識、話者分離、翻訳、個人情報マスキングを1回のリクエストで完結するパイプライン。

  • Complete turnkey pipeline: Transcription + Diarization + PII Redaction + Translation in one call
  • Saves up to 40% in latency and compute overhead vs chaining individual APIs
POST /v1/audio/pipelinecURL
curl -X POST https://api.brainiall.com/v1/audio/pipeline \
  -H "Authorization: Bearer $BRAINIALL_KEY" \
  -F "audio=@call_center.wav" \
  -F "features=transcribe,diarize,redact_pii,translate" \
  -F "target_lang=es"

今すぐ音声APIの利用を開始

リアルタイムの音声・音響インテリジェンスを本番システムに導入しましょう。

無料トライアルを開始
音声・音響インテリジェンスAPI | Brainiall | Brainiall