generated: '2026-07-21'
method: searched
source: >-
  https://voiceitt-si-api.readme.io/ — the cross-cutting request/response
  semantics documented in the Voiceitt HTTP and WebSockets getting-started
  guides, plus what the harvested OpenAPI declares.
description: >-
  How the Voiceitt API behaves across operations: JWT session authentication
  with refresh rotation, multipart transcription requests with a JSON options
  file, Socket.IO event streaming with partial/final recognition results, and
  strict PCM audio input requirements. The surface is small (auth +
  transcription) — there is no pagination, no idempotency-key mechanism, and
  no documented rate-limit signaling.
base_url: https://api2.voiceitt.com
websocket_url: https://web.voiceitt.com/socket.io
api_style: REST over HTTPS (multipart/form-data + JSON) plus Socket.IO WebSockets streaming

authentication:
  scheme: Bearer JWT obtained via App ID + API key login; refresh-token rotation
  modes:
    speaker_independent: POST /v1/auth/login/user_id (App ID + API key, optional user_id)
    personalized: POST /v1/auth/login/email (end user enrolled at https://web.voiceitt.com/)
  detail: authentication/voiceitt-authentication.yml

idempotency:
  supported: false
  notes: No idempotency-key header or replay semantics documented.

pagination:
  style: none
  notes: No list endpoints in the public API surface.

request_options:
  mechanism: >-
    Transcription behavior is controlled by an options JSON document —
    uploaded as the options part of the multipart /v1/rec/transcribe request,
    or sent via the set_options WebSocket event (repeatable at any time after
    connecting).
  fields:
  - name: run_itn
    type: boolean
    description: Inverse text normalization of the transcript.
  - name: run_capitalization
    type: boolean
    description: Capitalization processing.
  - name: run_spoken_command_conversion
    type: boolean
    description: Spoken command / punctuation conversion.
  - name: filter_profanities
    type: boolean
    description: Profanity filtering.
  - name: save_audio
    type: boolean
    description: Whether the server retains the uploaded audio (HTTP transcribe options example).
  - name: streaming_vad_min_silence_length
    type: float
    description: Silence duration required for a streaming speech segment to be considered complete (WebSockets).

audio_input:
  pcm:
    channels: 1
    sample_rate_hz: 16000
    sample_types: [int16, int32, float32, float64]
    recommended_chunk: 3200 samples per stream_audio_samples message
  compressed:
    formats: [mp4, ogg]
    notes: Sent via stream_compressed_audio with a MIME type; MediaRecorder recommended in browsers for lower distortion.

request_tracing:
  mechanism: request_id
  notes: >-
    recognize_audio_samples replies immediately with a request_received event
    carrying request_id; errors reference the same request_id.

versioning:
  scheme: uri-path
  current: v1
  detail: lifecycle/voiceitt-lifecycle.yml

error_envelope:
  http: Status codes with plain descriptions (e.g. 403 Invalid App ID or API key); no application/problem+json.
  websocket: error event with message and optional request_id.
  detail: errors/voiceitt-problem-types.yml

rate_limits:
  documented: false
  notes: No rate-limit headers or quotas documented in the public docs.