## Soniox docs summary:

Table of contents (links to pages)
- Community and support — https://soniox.com/docs/community-and-support
- Introduction — https://soniox.com/docs/
- Get started — https://soniox.com/docs/stt/get-started
- Models — https://soniox.com/docs/stt/models
- Security and privacy — https://soniox.com/docs/stt/security-and-privacy
- Web SDK — https://soniox.com/docs/stt/SDKs/web-sdk
- API reference — https://soniox.com/docs/stt/api-reference
- WebSocket API — https://soniox.com/docs/stt/api-reference/websocket-api
- Async transcription — https://soniox.com/docs/stt/async/async-transcription
- Async translation — https://soniox.com/docs/stt/async/async-translation
- Error handling (async) — https://soniox.com/docs/stt/async/error-handling
- Limits & quotas (async) — https://soniox.com/docs/stt/async/limits-and-quotas
- Webhooks — https://soniox.com/docs/stt/async/webhooks
- Confidence scores — https://soniox.com/docs/stt/concepts/confidence-scores
- Context — https://soniox.com/docs/stt/concepts/context
- Language hints — https://soniox.com/docs/stt/concepts/language-hints
- Language identification — https://soniox.com/docs/stt/concepts/language-identification
- Speaker diarization — https://soniox.com/docs/stt/concepts/speaker-diarization
- Supported languages — https://soniox.com/docs/stt/concepts/supported-languages
- Timestamps — https://soniox.com/docs/stt/concepts/timestamps
- Soniox Live — https://soniox.com/docs/stt/demo-apps/soniox-live
- Best practices — https://soniox.com/docs/stt/guides/best-practices
- Direct stream — https://soniox.com/docs/stt/guides/direct-stream
- Proxy stream — https://soniox.com/docs/stt/guides/proxy-stream
- Community integrations — https://soniox.com/docs/stt/integrations/community-integrations
- LiveKit — https://soniox.com/docs/stt/integrations/livekit
- Pipecat — https://soniox.com/docs/stt/integrations/pipecat
- Twilio — https://soniox.com/docs/stt/integrations/twilio
- Connection keepalive — https://soniox.com/docs/stt/rt/connection-keepalive
- Endpoint detection — https://soniox.com/docs/stt/rt/endpoint-detection
- Error handling (real-time) — https://soniox.com/docs/stt/rt/error-handling
- Limits & quotas (real-time) — https://soniox.com/docs/stt/rt/limits-and-quotas
- Manual finalization — https://soniox.com/docs/stt/rt/manual-finalization
- Real-time transcription — https://soniox.com/docs/stt/rt/real-time-transcription
- Real-time translation — https://soniox.com/docs/stt/rt/real-time-translation
- Create temporary API key — https://soniox.com/docs/stt/api-reference/auth/create_temporary_api_key
- Files API: Delete file — https://soniox.com/docs/stt/api-reference/files/delete_file
- Files API: Get file — https://soniox.com/docs/stt/api-reference/files/get_file
- Files API: Get file URL — https://soniox.com/docs/stt/api-reference/files/get_file_url
- Files API: Get files — https://soniox.com/docs/stt/api-reference/files/get_files
- Files API: Upload file — https://soniox.com/docs/stt/api-reference/files/upload_file
- Get models — https://soniox.com/docs/stt/api-reference/models/get_models
- Create transcription — https://soniox.com/docs/stt/api-reference/transcriptions/create_transcription
- Delete transcription — https://soniox.com/docs/stt/api-reference/transcriptions/delete_transcription
- Get transcription — https://soniox.com/docs/stt/api-reference/transcriptions/get_transcription
- Get transcription transcript — https://soniox.com/docs/stt/api-reference/transcriptions/get_transcription_transcript
- Get transcriptions — https://soniox.com/docs/stt/api-reference/transcriptions/get_transcriptions

Community and support
- Channels: GitHub for issues and code (https://github.com/soniox); Discord for fast support and discussion (https://discord.gg/rWfnk9uM5j).
- Docs MCP server: add remote config to your dev tool to query docs (example npx mcp-remote URL snippet).
- Website: https://soniox.com for product, billing, Console.

Introduction
- Soniox offers production-ready Speech-to-Text (STT) + translation APIs (async files + real-time streaming), speaker diarization, language ID, context/customization, timestamps, confidence scores.
- Console: signup and manage API keys, usage, billing at https://console.soniox.com.

Get started (STT)
- Quick flow:
  1. Create Soniox account and obtain API key (Console).
  2. Clone examples repository (https://github.com/soniox/soniox_examples) and run samples.
  3. Try real-time and async examples (Python/Node) included in repo.
- Environment example:
  export SONIOX_API_KEY=<YOUR_API_KEY>
- Example commands (Python):
  python soniox_realtime.py --audio_path ../assets/coffee_shop.mp3
  python soniox_async.py --audio_url "https://soniox.com/media/examples/coffee_shop.mp3"

Models
- Active models:
  - stt-rt-v3 — Real-time (active)
  - stt-async-v3 — Async (active)
- Aliases for stability: stt-rt-v3-preview → stt-rt-v3, stt-async-preview-v1 → stt-async-v3, etc.
- v3 model improvements: up to 5 hours audio, better multilingual switching, higher accuracy, improved diarization and translation.
- To upgrade: replace model name field: {"model": "stt-rt-v3"} or {"model": "stt-async-v3"}.

Security and privacy
- Certifications: SOC 2 Type 2, GDPR, HIPAA support. Request compliance docs via support@soniox.com.
- Data handling:
  - No model training on customer audio.
  - No retention unless user opts for storage (e.g., using Files API).
  - Stored data isolated per account; deletable via Console/API.
- Logging: minimal; logs never contain raw audio/transcripts.
- Encryption: TLS 1.2+ in transit, industry-standard at rest; access restricted by API keys.

Web SDK (speech-to-text-web)
- Purpose: official JS/TS client to use Real-Time API from browsers. Captures mic audio, streams to Soniox, returns partial/final tokens.
- Install: npm install @soniox/speech-to-text-web (or yarn/pnpm). CDN import available.
- Quickstart (essentials):
  - Construct SonioxClient({ apiKey, bufferQueueSize?, callbacks... })
  - start({ model, languageHints, context, enableSpeakerDiarization, enableLanguageIdentification, enableEndpointDetection, translation, stream, audioConstraints, mediaRecorderOptions, onPartialResult, onFinished, onError })
  - stop() — graceful; cancel() — immediate.
- Translation config examples:
  - One-way: { type: "one_way", target_language: "en" }
  - Two-way: { type: "two_way", language_a: "en", language_b: "es" }
- Temporary API keys: pass apiKey as async function that fetches a temporary key from your server; audio buffered until key arrives.
- Custom MediaStream: pass stream option (you manage lifecycle).

API reference (overview)
- Base: https://api.soniox.com/v1
- Major groups:
  - Auth API: POST /v1/auth/temporary-api-key
  - Files API: POST /v1/files (upload), GET /v1/files, GET /v1/files/{file_id}, GET /v1/files/{file_id}/url, DELETE /v1/files/{file_id}
  - Models API: GET /v1/models
  - Transcriptions API: POST /v1/transcriptions, GET /v1/transcriptions, GET /v1/transcriptions/{id}, GET /v1/transcriptions/{id}/transcript, DELETE /v1/transcriptions/{id}
- WebSocket API (real-time): wss://stt-rt.soniox.com/transcribe-websocket

WebSocket API (real-time)
- Endpoint: wss://stt-rt.soniox.com/transcribe-websocket
- Session start: send initial JSON config text message (before sending binary audio frames).
- Example start message (concise):

{
  "api_key": "<API_KEY|TEMP_KEY>",
  "model": "stt-rt-v3",
  "audio_format": "auto",
  "language_hints": ["en","es"],
  "context": {
    "general": [{"key":"domain","value":"Healthcare"}, ...],
    "text": "Long background text ...",
    "terms": ["Celebrex", "Zyrtec"],
    "translation_terms":[{"source":"Mr. Smith","target":"Sr. Smith"}]
  },
  "enable_speaker_diarization": true,
  "enable_language_identification": true,
  "enable_endpoint_detection": true,
  "translation": { "type":"two_way", "language_a":"en", "language_b":"es" },
  "client_reference_id": "optional-string"
}

- After config, stream audio as binary WebSocket frames (or send empty frame/text to end audio).
- Controls (JSON control messages during session):
  - Keepalive: {"type":"keepalive"} — send at least every 20s when audio paused.
  - Finalize: {"type":"finalize"} or with trailing_silence_ms: {"type":"finalize","trailing_silence_ms":300}.
  - Manual control messages must be sent as JSON text frames.

Audio formats
- audio_format: "auto" recommended for browser/containers.
- Raw audio needs audio_format + sample_rate + num_channels.
- Supported raw encodings: pcm_s16le, pcm_s32le, pcm_f32le, pcm_s8, pcm_u8, pcm_s24, mulaw, alaw, etc. (see docs for full list).
- Each stream supports up to 300 minutes (5 hours) audio.

Responses (JSON)
- Typical token response:

{
  "tokens": [
    { "text":"Hello", "start_ms":600, "end_ms":760, "confidence":0.97, "is_final":true, "speaker":"1", "language":"en" }
  ],
  "final_audio_proc_ms": 760,
  "total_audio_proc_ms": 880
}

- Finished response (session complete):

{
  "tokens": [],
  "final_audio_proc_ms": 1560,
  "total_audio_proc_ms": 1680,
  "finished": true
}

- Error response (server returns then closes connection):

{
  "tokens": [],
  "error_code": 503,
  "error_message": "Cannot continue request (code N). Please restart the request..."
}

Token fields (per-token)
- text (string)
- start_ms, end_ms (numbers, ms) — for spoken/original tokens (not for translated tokens)
- confidence (0.0–1.0)
- is_final (boolean) — finalized tokens are stable and never change
- speaker (string) — present if diarization enabled
- translation_status — "none" | "original" | "translation"
- language (ISO code) — language of token.text
- source_language — present for translation tokens (source original language)

Translation (real-time)
- Modes:
  - one_way: translate all spoken languages into single target language (translation.target_language required)
  - two_way: translate between language_a and language_b (language_a/language_b required)
- Token stream: transcription tokens (translation_status: "original" or "none") appear first with timestamps; translation tokens (translation_status: "translation") follow and do not include timestamps.
- Translations stream mid-sentence (low-latency) and may not align 1:1 with tokens.
- Example token fields for translated token: {"text":"Bonjour", "translation_status":"translation","language":"fr","source_language":"en"}

Endpoint detection & manual finalization
- enable_endpoint_detection: model detects end of utterance and emits <end> token (is_final true). Use to trigger downstream actions.
- Manual finalize: send {"type":"finalize"}; server finalizes pending tokens and emits {"text":"<fin>", "is_final":true}.
- For trailing silence optimization: {"type":"finalize","trailing_silence_ms":300}.

Connection keepalive
- Send {"type":"keepalive"} at least every 20s when no audio to prevent session timeout. Charge accrues for stream duration.

Errors and codes (summary)
- 400 Bad Request — malformed request or invalid params (missing audio_format for PCM, invalid model, context too long, no audio, audio decode errors, received too much audio, etc.)
- 401 Unauthorized — missing/invalid API key or expired temporary key
- 402 Payment required — billing/usage limits reached
- 408 Request timeout — inactivity, slow input, start timeout, audio decode timeout
- 429 Too many requests — rate/usage limits exceeded, concurrent requests
- 500 Internal server error — retry, contact support if persistent
- 503 Service unavailable — request cannot continue ("Cannot continue request (code N)"), client should restart the request

Async transcription (files)
- Use the Files API to upload local audio then create transcription referencing file_id; or pass audio_url for public URLs.
- Submit a transcription job: POST /v1/transcriptions with JSON body containing:
  - model, audio_url or file_id (one required), language_hints, context, enable_speaker_diarization (bool), enable_language_identification (bool), translation (one_way/two_way), client_reference_id, webhook_url, webhook_auth_header_name, webhook_auth_header_value.
- Polling: GET /v1/transcriptions/{transcription_id} — returns status (pending, processing, completed, error) and details.
- Retrieve full transcript: GET /v1/transcriptions/{id}/transcript — returns tokens (same token structure as real-time for completed jobs).
- Delete: DELETE /v1/transcriptions/{id} — permanently deletes transcript and associated files (cannot delete while processing).
- Example (Python flow summary):
  1. Optionally upload: POST /v1/files (multipart/form-data file)
  2. Create transcription: POST /v1/transcriptions {model, file_id|audio_url, ...}
  3. Poll GET /v1/transcriptions/{id} until status completed or error
  4. GET /v1/transcriptions/{id}/transcript to get tokens/text
  5. Optionally DELETE transcription and DELETE file(s)

Files API (summary)
- POST /v1/files — upload file; returns id. Use multipart/form-data "file".
- GET /v1/files — list files (pagination via next_page_cursor)
- GET /v1/files/{file_id} — metadata for uploaded file
- GET /v1/files/{file_id}/url — temporary (1-hour) signed URL for download
- DELETE /v1/files/{file_id} — delete file permanently

Create temporary API key
- POST /v1/auth/temporary-api-key
- Purpose: issue short-lived client keys for browser direct connections to WebSocket API.
- Required example body:
  { "usage_type": "transcribe_websocket", "expires_in_seconds": 60 }
- Temporary keys should be created on server side (use your long-lived API key) and returned to clients.

Webhooks (async)
- Provide webhook_url in transcription creation. Soniox sends POST when transcription completes or fails:
  { "id": "<transcription_id>", "status": "completed" | "error" }
- Authentication: provide webhook_auth_header_name and webhook_auth_header_value during transcription creation — Soniox includes the header in deliveries.
- Retry: Soniox retries several times; if permanently failed you can fetch results by transcription ID. Add query params to your webhook URL for metadata (e.g., ?customer_id=123).

Confidence scores
- Every token includes confidence (0.0–1.0). Use it to flag uncertain words for review/UI or post-processing.

Context (customization)
- context object improves transcription/translation: supports up to four sections:
  - general: array of {key, value} pairs (domain, topic, participants). Recommend ≤10 key-value pairs.
  - text: long unstructured text (background docs, meeting notes).
  - terms: array of strings (custom vocabulary, uncommon terms).
  - translation_terms: array of {source, target} for custom translations/preservations.
- Size limit: ~8,000 tokens (~10,000 chars). Too large → API error. Use to bias recognition and translations.

Language hints
- language_hints: array of ISO language codes (e.g., ["en","es"]). Hint biases recognition toward listed languages but does not restrict detection.
- Useful when likely languages known and for rare/similar languages.

Language identification
- enable_language_identification: true → each token includes language field.
- Token-level labeling, but model aims for sentence-level coherence; in real-time initial misclassifications may be revised as more context arrives.

Speaker diarization
- enable_speaker_diarization: true → tokens include speaker labels (e.g., "1","2"). Up to 15 speakers supported.
- Async diarization generally more accurate than real-time due to full-context processing.

Timestamps
- start_ms and end_ms are provided for tokens (in ms). Always included by default for spoken tokens (not for translated tokens).

Supported languages
- 60+ languages supported for transcription and translation. (List included in docs; ISO code examples: en, es, de, fr, zh, ja, ko, etc.)
- Programmatic retrieval: GET /v1/models returns supported languages metadata.

Real-time transcription (behavior)
- Token stream: non-final tokens (is_final:false) for immediate feedback; final tokens (is_final:true) eventually emitted and stable.
- Audio progress metrics per response:
  - final_audio_proc_ms — audio processed into final tokens
  - total_audio_proc_ms — audio processed into final + non-final tokens
- Ways to obtain final tokens sooner: endpoint detection or manual finalize.

Real-time translation (behavior)
- Transcription always produced; translations emitted as separate tokens (translation_status: "translation") following original tokens.
- Timestamps present for original tokens only.
- One-way and two-way modes supported (see translation object fields).

Best practices
- Provide language_hints and context when available for better accuracy.
- Use endpoint_detection or finalize to control latency vs accuracy tradeoffs.
- For real-time low-latency UIs: display non-final tokens live; act on final tokens or on <end>/<fin> markers.
- For speaker attribution and difficult audio, prefer async processing for higher diarization accuracy.

Direct stream (browser)
- Architecture: client obtains temporary API key from your server, uses RecordTranscribe / SonioxClient to connect directly to WebSocket endpoint and stream microphone audio.
- Temporary API key creation: server-side POST /v1/auth/temporary-api-key with usage_type "transcribe_websocket" and short expires_in_seconds (ex: 60).
- Client flow (minimal):
  1. fetch('/temporary-api-key') on server
  2. new RecordTranscribe({ apiKey: temporaryApiKey })
  3. start({ model:"stt-rt-v3", languageHints:["en"], onPartialResult, onFinished })
  4. stop() to finalize or cancel()
- HTML example exists in repo (index.html) showing obtaining temp key and using RecordTranscribe.

Proxy stream
- Pattern: client streams audio to your proxy WebSocket; proxy connects to Soniox WebSocket with your long-lived API key, forwards audio and relays responses.
- Useful when server-side inspection/transformation/storage required. Direct stream recommended for lowest latency.

Examples & SDKs / Integrations
- LiveKit plugin, Pipecat integration, Twilio streaming examples available in GitHub repos referenced in docs.
- Twilio: stream call audio (Start Stream TwiML) to your websocket server which forwards to Soniox; examples & repo provided.
- Pipecat: SonioxSTTService integrates with Pipecat pipelines; supports language_hints, context, vad_force_turn_endpoint options.

Limits & quotas
- Files (async):
  - Total file storage: default 10 GB
  - Uploaded files: default 1,000
  - File duration: 300 minutes (5 hours) — fixed
- Transcription counts:
  - Pending transcriptions limit: 100
  - Total transcriptions (pending + completed + failed): 2,000
- WebSocket real-time:
  - Requests per minute: 100
  - Concurrent connections per project/org: 10
  - Stream duration: 300 minutes per session
- To increase most limits use Soniox Console (except audio duration).

Error handling (async & real-time)
- Async:
  - File upload errors: check duration ≤ 300 minutes, storage/quota; recover by deleting files or requesting higher limits.
  - Transcription creation: respect pending/total transcription limits; delete older results if needed.
  - Webhook failures: Soniox retries; if permanently failed, fetch transcription result via API using transcription ID.
- Real-time:
  - Errors are returned in JSON and connection closed. Log error_code and error_message. For "Cannot continue request" 503, start a fresh session.
  - If audio input is too slow or missing start message within timeouts, server closes connection (408).

Representative code snippets (condensed)

1) WebSocket start config (JSON)
{
  "api_key":"<API_KEY>",
  "model":"stt-rt-v3",
  "audio_format":"auto",
  "language_hints":["en","es"],
  "context":{ "terms":["TermA","TermB"], "text":"Background..." },
  "enable_speaker_diarization": true,
  "enable_language_identification": true,
  "enable_endpoint_detection": true
}

2) End-of-audio signal
- Send an empty WebSocket frame/string to indicate end-of-audio; server will return finished response and close connection.

3) Create async transcription (curl-style)
POST https://api.soniox.com/v1/transcriptions
Headers: Authorization: Bearer $SONIOX_API_KEY
Body (JSON):
{
  "model":"stt-async-v3",
  "audio_url":"https://example.com/audio.mp3",
  "language_hints":["en","es"],
  "enable_speaker_diarization": true,
  "context": { "terms":["BrandX"], "text":"Meeting notes..." },
  "translation": { "type":"one_way", "target_language":"es" },
  "client_reference_id":"MyRef123",
  "webhook_url":"https://example.com/webhook",
  "webhook_auth_header_name":"Authorization",
  "webhook_auth_header_value":"Bearer secret"
}

4) Temporary API key creation (server-side)
POST https://api.soniox.com/v1/auth/temporary-api-key
Headers: Authorization: Bearer <SONIOX_API_KEY>, Content-Type: application/json
Body:
{ "usage_type":"transcribe_websocket", "expires_in_seconds":60 }
Response: { "api_key":"<temporary_key>", "expires_in_seconds":60, ... }

5) Minimal Web SDK (client) pattern (pseudo):
const client = new SonioxClient({ apiKey: async () => fetch('/tmp-key').then(r=>r.json()).then(j=>j.api_key) });
client.start({
  model:'stt-rt-v3',
  languageHints:['en'],
  context:'Product names, jargon...',
  enableSpeakerDiarization:true,
  onPartialResult: r => console.log(r.tokens),
  onFinished: ()=>console.log('done')
});

6) Python real-time snippet (pseudo):
ws = connect("wss://stt-rt.soniox.com/transcribe-websocket")
ws.send(json.dumps(config))
# stream binary chunks
ws.send(binary_chunk)  # repeat
# end
ws.send("")

Repository examples
- Full runnable samples in the official examples repo: https://github.com/soniox/soniox_examples — includes python/node real-time and async samples, direct stream, proxy stream, Soniox Live demo.

Where to find more
- API OpenAPI spec and full endpoint schemas available in the API Reference pages (links in TOC).
- Example repos and integration guides contain full runnable code.

This summary preserves key pages, endpoints, required parameters, response formats, common control messages, error handling, limits, SDK usage and representative code snippets to implement both async and real-time transcription/translation flows.

## Soniox website summary:

Helping startups and enterprises ship real world voice apps
- Production-ready real-time voice AI: transcription, translation, and understanding in 60+ languages. Low-latency, speaker-aware, works in noisy/overlapping real-world audio.

For developers
- Single API for real-time streaming and asynchronous (file) transcription, translation, speaker separation, language detection, domain hints, structured JSON.
- Docs & API reference: https://soniox.com/docs

For everyone
- Mobile app demonstrating live transcription, translation, summaries, speaker labels, export and privacy-first processing.
- iOS App: https://apps.apple.com/us/app/soniox/id1560199731
- Android App: https://play.google.com/store/apps/details?id=com.soniox.sonioxmobileapp

Explore the API
- Real-time token-level output (milliseconds), async file mode, unified stream for transcription+translation, domain/context hints, alphanumeric accuracy, endpoint detection, speaker diarization.
- Start: create account & generate API key (see docs)

Everything you need to build great voice apps (Soniox Speech-to-Text AI)
- One universal model for 60+ languages (real-time + async)
- Features: instant translation, language ID, speaker labels, timestamps, punctuation, structured output, custom terms/context
- Privacy & compliance: audio processed in memory by default; SOC 2 Type II, HIPAA-ready, GDPR-aligned

About us
- Mission: make speech AI universal — understand people through audio.
- Founded: 2020; breakthrough unsupervised training for speech recognition.
- Founding team:
  - Klemen Simonic — Founder, CEO
  - Ambroz Bizjak — Co-founder, Chief Architect
- Locations:
  - Soniox: 1045 Helm Ln, Foster City, CA 94404, United States
  - Soniox Europe: Cesta v Gorice 34B, 1000 Ljubljana, Slovenia

Benchmarks
- Independent multi-language benchmarks (60 languages, real-world YouTube datasets) show Soniox achieving top accuracy vs major providers. Benchmark report available on site.

Blog (highlights)
- Soniox v3 — Oct 21, 2025: major model update (stt-rt-v3, stt-async-v3)
- Free credits update — Oct 27, 2025
- Real-time speech translation launch — Jun 17, 2025
- Introducing Soniox app — May 6, 2025
- Full blog archive on site

Compare
- Soniox Compare: live side-by-side real-time tests vs other providers; open-source framework so you can reproduce tests.

Contact
- General / partnerships: info@soniox.com
- Technical / support: support@soniox.com
- Feedback / product: hello@soniox.com
- Community Discord: https://discord.gg/rWfnk9uM5j

Policies / Compliance
- Default: audio not stored or used to train models unless explicitly requested.
- Certifications: SOC 2 Type II, HIPAA-ready; GDPR compliance.
- To request data compliance documentation: support@soniox.com

Pricing (summary)
- Token-based pricing; simple hourly equivalents:
  - Approx. $0.12 / hour (real-time streaming)
  - Approx. $0.10 / hour (async/file)
- Mobile app: free tier (10 weekly credits); Soniox Pro $19.99/month (unlimited, priority processing)

Soniox App (mobile)
- Live transcription, translation, speaker labels, summaries, to-dos, emotion/tone insights, export.
- Privacy-first: audio processed in memory, not used to train models by default.
- Get apps:
  - iOS: https://apps.apple.com/us/app/soniox/id1560199731
  - Android: https://play.google.com/store/apps/details?id=com.soniox.sonioxmobileapp

Soniox API
- One API to transcribe, translate, detect language, separate speakers, and stream token-level output for fast voice apps.
- Use cases: voice agents, call centers, media captions, medical transcription, analytics, wearables, real-time translation.
- Docs: https://soniox.com/docs

Use cases (brief)
- Call centers & support automation: live multilingual transcription, summaries, structured output for CRM/QA.
- Real-time media transcription: captions, subtitles, speaker labels, publish-ready transcripts.
- Medical transcription: HIPAA-ready real-time clinical notes, custom medical vocab, speaker labels.
- Real-time speech analytics: sentiment, tone, compliance, live coaching.
- Translate speech instantly in 60+ languages: any-to-any streaming, mid-sentence switching.
- Build real-time voice agents: low-latency, structured transcripts for routing, triage, assistants.
- Power wearable apps: lightweight real-time speech for watches, glasses, earbuds; private processing.

Key links & contacts
- Docs: https://soniox.com/docs
- iOS app: https://apps.apple.com/us/app/soniox/id1560199731
- Android app: https://play.google.com/store/apps/details?id=com.soniox.sonioxmobileapp
- Discord: https://discord.gg/rWfnk9uM5j
- Email: info@soniox.com, support@soniox.com, hello@soniox.com

(End of site summary)