Name: NVIDIA Parakeet TDT 0.6B v3 Speech-to-Text API
Brand: magicapi
SKU: api-magicapi-audio-speech-to-text-dmtanner-parakeet-tdt-0-6b-v3
Availability: InStock
Author: API Market

Shashank Agarwal

# 🎙️ Speech-to-Text API with Timestamps (NVIDIA Parakeet TDT 0.6B v3)

## Overview
Parakeet TDT 0.6B v3 is a speech recognition (ASR) model created by NVIDIA. It transcribes speech to text and can return word-level timestamps.

Send an `audio_file` (wav, flac or mp3) or an `m3u8_url` playlist. Use `start_time` and `end_time` to transcribe only a section of an m3u8 stream. The response is a JSON string with the transcript in `text`, plus character, word and segment timestamps when `timestamps` is `true`.

## Features
- **Audio files**: send a wav, flac or mp3 file as `audio_file`.
- **Streams**: send an `m3u8_url` playlist with .m4s or .ts segments.
- **Time window**: `start_time` and `end_time` (seconds) limit transcription to one section of an m3u8 stream.
- **Timestamps**: set `timestamps` to `true` (default `false`) to get start and end times per word, per character group and per segment.
- **Punctuation and casing**: the example transcript comes back with capital letters, commas and a full stop.

## Use Cases
1. **Subtitles and captions**: turn word timestamps into subtitle cues.
2. **Searchable recordings**: index podcasts, meetings and lectures by the words spoken and the time they were spoken.
3. **Live stream clips**: transcribe a slice of an m3u8 stream by setting `start_time` and `end_time`.
4. **Voice notes**: convert short spoken notes into text for a database or chat.

## Visual Examples
### Example Input
- `audio_file`: a short read-speech clip (16.6 seconds of speech). [Listen to the input audio](https://blog.api.market/wp-content/uploads/2026/10/granite-speech-3-3-8b-input-api-market.wav)
- `timestamps`: `true`

### Example Output
The JSON string returned in `output`, decoded and shortened (the full response lists every timestep, character group and word):

```json
{
  "text": "After his nap, Timothy lazily stretched first one grey velvet foot, then another, strolled indolently to his plate, turning over the food, carefully selecting choice bits, nosing out that which he scorned upon the clean hearth.",
  "timestamps": {
    "timestep": [7, 11, 14],
    "char": [
      {"char": ["A"], "start_offset": 7, "end_offset": 11, "start": 0.56, "end": 0.88},
      {"char": ["fter"], "start_offset": 11, "end_offset": 14, "start": 0.88, "end": 1.12}
    ],
    "word": [
      {"word": "After", "start_offset": 7, "end_offset": 14, "start": 0.56, "end": 1.12},
      {"word": "his", "start_offset": 14, "end_offset": 17, "start": 1.12, "end": 1.36},
      {"word": "nap,", "start_offset": 17, "end_offset": 20, "start": 1.36, "end": 1.6}
    ],
    "segment": [
      {"segment": "After his nap, Timothy lazily stretched first one grey velvet foot, then another, strolled indolently to his plate, turning over the food, carefully selecting choice bits, nosing out that which he scorned upon the clean hearth.", "start_offset": 7, "end_offset": 208, "start": 0.56, "end": 16.64}
    ]
  }
}
```

The `segment` entry holds the whole sentence, spoken from 0.56 s to 16.64 s.

## Input Parameters
| Field | Type | Required | Default | Allowed values | Description |
| --- | --- | --- | --- | --- | --- |
| `audio_file` | string (URL) | No | none |  | Audio file URL (wav, flac or mp3). Send this or m3u8_url. |
| `m3u8_url` | string | No | none |  | URL of an m3u8 playlist (supports .m4s and .ts segments). Send this or audio_file. |
| `start_time` | number | No | none |  | Start time in seconds of the window to transcribe |
| `end_time` | number | No | none |  | End time in seconds of the window to transcribe |
| `timestamps` | boolean | No | `false` |  | Whether to include timestamps in the transcription |

## Pricing
- **Pay As You Go (PAYG) pricing**: you only pay for what you use.
- **Price**: $0.0022 per API unit.
- **1 API unit = 1 request with up to 2 seconds of processing.** A request that needs more processing time is billed one more unit for each started 2-second block, so more than 2 and up to 4 seconds of processing is 2 units ($0.0044), and 9 seconds is 5 units ($0.011). Longer audio takes longer to process.
- **Example cost**: the example request processed for 1.53 seconds, so it used 1 API unit and cost $0.0022.
- **Failed or cancelled predictions are not billed.**
- **GET requests to check prediction status** are free and don't count toward API unit usage.
- **No monthly subscription fee.**

## API Documentation

### Base URL
All API requests go to: `https://prod.api.market/api/v1/magicapi/audio-speech-to-text-dmtanner-parakeet-tdt-0-6b-v3`

### Authentication
Send your api.market API key in the `x-api-market-key` header:
```
x-api-market-key: YOUR_API_MARKET_KEY
```

### Asynchronous Processing
**IMPORTANT**: This API is asynchronous. The flow is:
1. Send a POST request to `/predictions`. A successful request returns **201 Created** with a prediction `id` and a `status` of `starting` or `processing`.
2. Poll GET `/predictions/{id}` with that ID.
3. Keep polling while `status` is `starting` or `processing`.
4. Stop when `status` is `succeeded` (read `output`), `failed` or `canceled` (read `error`).

A polling interval of 1 second and an overall timeout of 600 seconds (10 minutes) work well.

### Endpoints

#### Create a Prediction
```bash
curl -X POST "https://prod.api.market/api/v1/magicapi/audio-speech-to-text-dmtanner-parakeet-tdt-0-6b-v3/predictions" \
  -H "Content-Type: application/json" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY" \
  -d '{
    "version": "74e605e7e05d1a7f52a5a7bb5d741a8b5a087309bee88ebdd249f63835f8a90d",
    "input": {
      "audio_file": "https://blog.api.market/wp-content/uploads/2026/10/granite-speech-3-3-8b-input-api-market.wav",
      "timestamps": true
    }
  }'
```

Response (201 Created):
```json
{
  "id": "vm40904mmxrmr0d1mvtskwzgc8",
  "version": "74e605e7e05d1a7f52a5a7bb5d741a8b5a087309bee88ebdd249f63835f8a90d",
  "status": "starting",
  "created_at": "2026-11-04T14:27:54.663Z"
}
```

#### Check Prediction Status
```bash
curl -X GET "https://prod.api.market/api/v1/magicapi/audio-speech-to-text-dmtanner-parakeet-tdt-0-6b-v3/predictions/{id}" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY"
```

Response when the prediction has finished (200 OK):
```json
{
  "id": "vm40904mmxrmr0d1mvtskwzgc8",
  "status": "succeeded",
  "output": "{\"text\": \"After his nap, Timothy lazily stretched first one grey velvet foot, then another, strolled indolently to his plate, turning over the food, carefully selecting choice bits, nosing out that which he scorned upon the clean hearth.\", \"timestamps\": {\"timestep\": [7, 11, 14], \"char\": [{\"char\": [\"A\"], \"start_offset\": 7, \"end_offset\": 11, \"start\": 0.56, \"end\": 0.88}, {\"char\": [\"fter\"], \"start_offset\": 11, \"end_offset\": 14, \"start\": 0.88, \"end\": 1.12}], \"word\": [{\"word\": \"After\", \"start_offset\": 7, \"end_offset\": 14, \"start\": 0.56, \"end\": 1.12}, {\"word\": \"his\", \"start_offset\": 14, \"end_offset\": 17, \"start\": 1.12, \"end\": 1.36}, {\"word\": \"nap,\", \"start_offset\": 17, \"end_offset\": 20, \"start\": 1.36, \"end\": 1.6}], \"segment\": [{\"segment\": \"After his nap, Timothy lazily stretched first one grey velvet foot, then another, strolled indolently to his plate, turning over the food, carefully selecting choice bits, nosing out that which he scorned upon the clean hearth.\", \"start_offset\": 7, \"end_offset\": 208, \"start\": 0.56, \"end\": 16.64}]}}",
  "error": null,
  "created_at": "2026-11-04T14:27:54.663Z",
  "started_at": "2026-11-04T14:29:17.678016363Z",
  "completed_at": "2026-11-04T14:29:19.20670803Z"
}
```

`output` is a string that holds JSON. Parse it to read `text` (the transcript). With `timestamps` set to `true` it also has `timestamps`, with `timestep` (frame offsets), `char` (character groups), `word` and `segment` lists; each entry has `start` and `end` in seconds plus `start_offset` and `end_offset`. The response above is shortened; the real one lists every entry.

### Status Codes
- **201 Created**: prediction created (POST `/predictions`). This is the normal success response.
- **200 OK**: prediction details returned (GET `/predictions/{id}`).
- **400 Bad Request**: invalid input, for example a `start_time` that is not a number.
- **401 Unauthorized**: missing or invalid API key.
- **404 Not Found**: no prediction with that ID.
- **429 Too Many Requests**: rate limit exceeded.
- **500 Internal Server Error** or **503 Service Unavailable**: temporary server problem; retry later.

## Rate Limits
Requests are rate limited to keep usage fair. If you receive a 429 response, wait and retry with exponential backoff.

## Error Handling
- Check the HTTP status code of every request and read the response body for the error message.
- A prediction can also end with `status` set to `failed`; the `error` field then explains why. Failed predictions are not billed.
- Retry 429, 500 and 503 responses with exponential backoff. Fix the input before retrying a 400.

## Implementation Guide
### Example Code

#### JavaScript
```javascript
const apiKey = "YOUR_API_MARKET_KEY";
const baseUrl = "https://prod.api.market/api/v1/magicapi/audio-speech-to-text-dmtanner-parakeet-tdt-0-6b-v3";

async function createAndWaitForPrediction(input) {
  const response = await fetch(`${baseUrl}/predictions`, {
    method: "POST",
    headers: {
      "Content-Type": "application/json",
      "x-api-market-key": apiKey
    },
    body: JSON.stringify({ version: "74e605e7e05d1a7f52a5a7bb5d741a8b5a087309bee88ebdd249f63835f8a90d", input })
  });

  if (response.status !== 201) {
    throw new Error(`API error: ${response.status} ${await response.text()}`);
  }

  const prediction = await response.json();
  return pollForResults(prediction.id);
}

const delay = (ms) => new Promise((resolve) => setTimeout(resolve, ms));

async function pollForResults(predictionId) {
  const maxTimeout = 600000;
  const pollInterval = 1000;
  const startTime = Date.now();

  while (Date.now() - startTime < maxTimeout) {
    const response = await fetch(`${baseUrl}/predictions/${predictionId}`, {
      headers: { "x-api-market-key": apiKey }
    });

    if (!response.ok) {
      throw new Error(`API error: ${response.status} ${await response.text()}`);
    }

    const prediction = await response.json();

    if (prediction.status === "succeeded") {
      return prediction.output;
    }

    if (prediction.status === "failed" || prediction.status === "canceled") {
      throw new Error(`Prediction ${prediction.status}: ${prediction.error || "Unknown error"}`);
    }

    await delay(pollInterval);
  }

  throw new Error("Prediction timed out after 10 minutes");
}

createAndWaitForPrediction({
  "audio_file": "https://blog.api.market/wp-content/uploads/2026/10/granite-speech-3-3-8b-input-api-market.wav",
  "timestamps": true
}).then((output) => console.log(JSON.parse(output).text));
```

#### Python
```python
import time
import requests

API_KEY = "YOUR_API_MARKET_KEY"
BASE_URL = "https://prod.api.market/api/v1/magicapi/audio-speech-to-text-dmtanner-parakeet-tdt-0-6b-v3"


def create_and_wait_for_prediction(input_data):
    headers = {"Content-Type": "application/json", "x-api-market-key": API_KEY}
    response = requests.post(
        f"{BASE_URL}/predictions",
        headers=headers,
        json={"version": "74e605e7e05d1a7f52a5a7bb5d741a8b5a087309bee88ebdd249f63835f8a90d", "input": input_data},
    )

    if response.status_code != 201:
        raise Exception(f"API error: {response.status_code} {response.text}")

    return poll_for_results(response.json()["id"])


def poll_for_results(prediction_id):
    max_timeout = 600
    poll_interval = 1
    headers = {"x-api-market-key": API_KEY}
    start_time = time.time()

    while time.time() - start_time < max_timeout:
        response = requests.get(f"{BASE_URL}/predictions/{prediction_id}", headers=headers)

        if response.status_code != 200:
            raise Exception(f"API error: {response.status_code} {response.text}")

        prediction = response.json()

        if prediction["status"] == "succeeded":
            return prediction["output"]

        if prediction["status"] in ("failed", "canceled"):
            raise Exception(f"Prediction {prediction['status']}: {prediction.get('error') or 'Unknown error'}")

        time.sleep(poll_interval)

    raise Exception("Prediction timed out after 10 minutes")


import json

print(json.loads(create_and_wait_for_prediction({
    "audio_file": "https://blog.api.market/wp-content/uploads/2026/10/granite-speech-3-3-8b-input-api-market.wav",
    "timestamps": True
}))["text"])
```

## FAQ
- **Which input do I send?**
  - Send either `audio_file` (a wav, flac or mp3 URL) or `m3u8_url` (a playlist URL). Both are optional in the schema, but one of them is needed for a transcript.
- **How do I get word timings?**
  - Set `timestamps` to `true`. Parse `output` as JSON and read `timestamps.word`: each word has `start` and `end` in seconds.
- **Can I transcribe part of a stream?**
  - Yes. Send `m3u8_url` with `start_time` and `end_time` in seconds to process only that window.
- **Why is the output a string and not an object?**
  - The model returns its JSON result as text. Run `JSON.parse` (JavaScript) or `json.loads` (Python) on `output`.
- **How much does a long recording cost?**
  - Billing follows processing time: one unit ($0.0022) per started 2 seconds of processing. The example clip (16.6 seconds of speech) processed in 1.53 seconds and cost one unit.
- **What if my prediction fails?**
  - Failed predictions are not billed. Read the `error` field, check that the audio URL is reachable and in a supported format, and try again.

## Contact Us
If you have any questions, reach out to us through the chat on the bottom right of the page.


Transcribe audio files (wav, flac, mp3) or m3u8 streams with NVIDIA Parakeet TDT 0.6B v3, with optional word-level timestamps. $0.0022 per call, pay as you go.