Word Timestamps

[securigy]

A simple question:

Is it possible to obtain start and stop timestamp of every word during transcription? If yes, how?

[StephenHodgson]

Hi @securigy. I'm unsure what you mean exactly of word timestamps. Is that in the docs anywhere?

[securigy]

I see that original EleventLabs APIs do not have timestamp for every word, like Google or Azure do. However, Websockets endpoint provides that, unfortunately in Python only.

{
    "audio": "Y3VyaW91cyBtaW5kcyB0aGluayBhbGlrZSA6KQ==",
    "isFinal": false,
    "normalizedAlignment": {
        "char_start_times_ms": [0, 3, 7, 9, 11, 12, 13, 15, 17, 19, 21],
        "chars_durations_ms": [3, 4, 2, 2, 1, 1, 2, 2, 2, 2, 3]
        "chars": ["H", "e", "l", "l", "o", " ", "w", "o", "r", "l", "d"]
    },
    "alignment": {
        "char_start_times_ms": [0, 3, 7, 9, 11, 12, 13, 15, 17, 19, 21],
        "chars_durations_ms": [3, 4, 2, 2, 1, 1, 2, 2, 2, 2, 3]
        "chars": ["H", "e", "l", "l", "o", " ", "w", "o", "r", "l", "d"]
    }
}

[StephenHodgson]

This is specifically for streaming socket, which isn't supported in my library.

[StephenHodgson]

I've been mulling over supporting web sockets but isn't a priority for me atm