---
title: "Run a phoneme-recognition pass on a recording. Provide either a previously-uploaded fileId (preferred for performance…"
method: POST
path: "/api/analyze-phonemes-live"
tags: ["Phoneme & Classification"]
---

# Run a phoneme-recognition pass on a recording. Provide either a previously-uploaded fileId (preferred for performance…

`POST /api/analyze-phonemes-live`

## Request

- Base URL: `https://platform.vocametrix.com`
- URL: `https://platform.vocametrix.com/api/analyze-phonemes-live`
- Auth: API key in header `X-API-Key`

## Request body

- object
  - `fileId` string: Optional (use this OR blobUrl). fileId from a prior /api/assignFileId upload.
  - `blobUrl` string: Optional (use this OR fileId). HTTPS URL of the audio file. Will be downloaded server-side. If the host is `vocametrixstorageaccount.blob.core.windows.net` and `keepBlob` is false, the blob is deleted after processing.
  - `referenceWord` string: Optional. Currently accepted by the API but not used by the JavaScript layer (passed through to the Python script which may use it).
  - `language` string: Optional, default "fr-FR". Accepted: "fr-FR" | "et-EE".
  - `model` string: Optional, French only. `short-words-champion` (isolated short words, recommended), `logatome-champion` (logatomes / nonsense syllables), or omit for the CNAM wav2vec2-french-phonemizer-v2 baseline (continuous speech). Must be a known alias (server-side `isKnownAlias`) — an unknown alias returns 400. Resolved to a model blob URL internally, never exposed to the client.
  - `keepBlob` string: Optional, default false. If false and `blobUrl` was used, the source blob is deleted after success (only when hosted on the Vocametrix storage account).

## Response `200`

Successful response

- object
  - `success` boolean: Boolean — true when analysis completed (including the empty-result "Silence detected" case).
  - `status` string: "completed" on success.
  - `message` string: Descriptive message, e.g. "Analysis completed successfully" or "Silence detected".
  - `language` string: Echoed language code processed.
  - `transcription` string: Space-separated sequence of detected phonemes (empty string if silence).
  - `phoneme_count` number: Total number of phonemes detected.
  - `phonemes` object[]: Array of unique phoneme labels, in order of first appearance (duplicates removed). Use `transcription` or `phoneme_timings` for the full sequence.
  - `phoneme_timings` object[]: Array of `{ phoneme, start, end, duration, confidence, second_phoneme, second_confidence }` — one entry per phoneme. `second_phoneme`/`second_confidence` (French champion models only) is the runner-up at that phoneme's most confident frame, or null if unavailable.
  - `audio_info` string: `{ duration_seconds, sample_rate }` — sample_rate is always 16000.

## Other responses

- `400`: Bad request — missing or invalid parameter
- `401`: Unauthorized — missing or invalid API key
- `429`: Rate limit exceeded
- `500`: Server error

---

[API](https://skmtc.dev/vocametrix/apis/vocametrix-api.md) · [All operations](https://skmtc.dev/vocametrix/apis/vocametrix-api/llms.txt) · [OpenAPI document](https://skmtc.dev/vocametrix/apis/vocametrix-api/revisions/8a54d01bcc81?raw)