Name: IBM Granite Speech 3.3 8B API: Speech to Text
Brand: magicapi
SKU: api-magicapi-automatic-speech-recognition-translation-ibm-granite-granite-speech-3-3-8b
Availability: InStock
Author: API Market

Shashank Agarwal

# 🎙️ Speech Recognition and Translation API (IBM Granite Speech 3.3 8B)

## Overview
Granite Speech 3.3 8B is a speech-language model from IBM. Its developer describes it as a compact and efficient model designed for automatic speech recognition (ASR) and automatic speech translation (AST).

Send one or more `audio` URLs together with a text `prompt` that says what to do, for example "Transcribe the speech into written form." The model answers in text. The output is a list of text chunks; join them to get the full answer.

## Features
- **Speech recognition**: transcribe spoken audio into text with a prompt such as "Transcribe the speech into written form."
- **Speech translation**: its developer describes the model as also built for speech translation; ask for it in the `prompt`.
- **Audio list**: `audio` takes an array of audio file URLs.
- **System prompt and chat template**: optional `system_prompt` and `chat_template` to change how the prompt is formatted.
- **Generation controls**: `max_tokens` (default 512), `min_tokens`, `temperature`, `top_p`, `top_k`, `presence_penalty`, `frequency_penalty`, `stop_sequences` and `seed`.
- **Pay for output only**: only output tokens are billed; input audio and prompt tokens are free.

## Use Cases
1. **Transcripts**: turn recorded speech into written text.
2. **Speech translation**: ask for the speech in another language in the `prompt`.
3. **Voice notes to text**: convert short voice messages into text for search or tickets.
4. **Captions**: produce the text of a clip as a starting point for captions.

## Visual Examples
### Example Input
[Listen to the example input audio](https://blog.api.market/wp-content/uploads/2026/10/granite-speech-3-3-8b-input-api-market.wav)

Prompt: "Transcribe the speech into written form."

Settings: `max_tokens` 512, `min_tokens` 0, `temperature` 0.6, `top_p` 0.9, `top_k` 50, `presence_penalty` 0, `frequency_penalty` 0.

### Example Output
The model returned a list of text chunks. Joined, they read:

> After his nap, Timothy leisurely stretched, first one gray velvet foot, then the other. He strolled indolently to his plate, turning over the food, carefully selecting choice bits. He sniffed out what he scorned upon the clean clean hearth.

## Input Parameters
| Field | Type | Required | Default | Allowed values | Description |
| --- | --- | --- | --- | --- | --- |
| `prompt` | string | No | `""` |  | User prompt to send to the model, for example "Transcribe the speech into written form." |
| `audio` | array of strings (URLs) | No | none |  | Audio inputs for the model |
| `system_prompt` | string | No | none |  | System prompt to send to the model. The chat template provides a good default |
| `min_tokens` | integer | No | `0` |  | The minimum number of tokens the model should generate as output |
| `max_tokens` | integer | No | `512` |  | The maximum number of tokens the model should generate as output |
| `temperature` | number | No | `0.6` |  | The value used to modulate the next token probabilities |
| `top_p` | number | No | `0.9` |  | A probability threshold for generating the output. If below 1.0, only the top tokens with cumulative probability at or above `top_p` are kept (nucleus filtering) |
| `top_k` | integer | No | `50` |  | The number of highest probability tokens to consider. If above 0, only the top k tokens are kept (top-k filtering) |
| `presence_penalty` | number | No | `0` |  | Presence penalty |
| `frequency_penalty` | number | No | `0` |  | Frequency penalty |
| `stop_sequences` | string | No | none |  | A comma-separated list of sequences to stop generation at. For example, `<end>,<stop>` stops generation at the first `<end>` or `<stop>` |
| `chat_template` | string | No | none |  | A template to format the prompt with. If not provided, the default prompt template is used |
| `seed` | integer | No | none |  | Random seed. Leave blank to randomize the seed |

## Pricing
- **Pay As You Go (PAYG) pricing**: you only pay for what you use.
- **Price**: $0.001 per API unit.
- **1 API unit = 100 weighted tokens; an output token counts 1, an input token counts 0.** Each request is rounded up to a whole unit.
- **Per million tokens**: input tokens $0 (audio and prompt are free), output tokens $10.
- The example request above used 237 input tokens and 64 output tokens: 237 x 0 + 64 x 1 = 64 weighted tokens, which rounds up to 1 unit, so it cost $0.001.
- With the default `max_tokens` of 512, one request costs at most 6 units ($0.006).
- **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/automatic-speech-recognition-translation-ibm-granite-granite-speech-3-3-8b`

### 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/automatic-speech-recognition-translation-ibm-granite-granite-speech-3-3-8b/predictions" \
  -H "Content-Type: application/json" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY" \
  -d '{
    "version": "688e7a943167401c310f0975cb68f1a35e0bddc3b65f60bde89c37860e07edf1",
    "input": {
      "audio": [
        "https://blog.api.market/wp-content/uploads/2026/10/granite-speech-3-3-8b-input-api-market.wav"
      ],
      "top_k": 50,
      "top_p": 0.9,
      "prompt": "Transcribe the speech into written form.",
      "max_tokens": 512,
      "min_tokens": 0,
      "temperature": 0.6,
      "presence_penalty": 0,
      "frequency_penalty": 0
    }
  }'
```

Response (201 Created):
```json
{
  "id": "nds5bpwa1srmw0d1d8g9qtbnbg",
  "version": "688e7a943167401c310f0975cb68f1a35e0bddc3b65f60bde89c37860e07edf1",
  "status": "starting",
  "created_at": "2026-10-23T18:58:00.169Z"
}
```

#### Check Prediction Status
```bash
curl -X GET "https://prod.api.market/api/v1/magicapi/automatic-speech-recognition-translation-ibm-granite-granite-speech-3-3-8b/predictions/{id}" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY"
```

Response when the prediction has finished (200 OK):
```json
{
  "id": "nds5bpwa1srmw0d1d8g9qtbnbg",
  "status": "succeeded",
  "output": [
    "After", " his", " nap", ",", " Tim", "oth", "y", " le", "is", "ure", "ly", " st", "ret", "ched", ",",
    " first", " one", " gray", " vel", "vet", " foot", ",", " then", " the", " other", ".",
    " He", " st", "rolled", " ind", "ol", "ently", " to", " his", " plate", ",", " turning", " over", " the", " food", ",",
    " carefully", " selecting", " choice", " bits", ".",
    " He", " sn", "iff", "ed", " out", " what", " he", " sc", "orn", "ed", " upon", " the", " clean", " clean", " he", "arth", ".", ""
  ],
  "error": null,
  "created_at": "2026-10-23T18:58:00.169Z",
  "started_at": "2026-10-23T18:58:01.169Z",
  "completed_at": "2026-10-23T18:58:04.169Z"
}
```

`output` is an array of text chunks. Join them without a separator to get the full answer.

### 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 `audio` that is not an array of URLs.
- **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
All plans allow 10 requests per second. Requests above that return `429 Too Many Requests`; back off before retrying and honour any `Retry-After` header. Polling a prediction with GET is free but still counts toward the per-second limit, so poll every 1 to 5 seconds rather than in a tight loop.

## 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/automatic-speech-recognition-translation-ibm-granite-granite-speech-3-3-8b";

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: "688e7a943167401c310f0975cb68f1a35e0bddc3b65f60bde89c37860e07edf1", 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.join("");
    }

    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": [
    "https://blog.api.market/wp-content/uploads/2026/10/granite-speech-3-3-8b-input-api-market.wav"
  ],
  "top_k": 50,
  "top_p": 0.9,
  "prompt": "Transcribe the speech into written form.",
  "max_tokens": 512,
  "min_tokens": 0,
  "temperature": 0.6,
  "presence_penalty": 0,
  "frequency_penalty": 0
}).then((text) => console.log(text));
```

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

API_KEY = "YOUR_API_MARKET_KEY"
BASE_URL = "https://prod.api.market/api/v1/magicapi/automatic-speech-recognition-translation-ibm-granite-granite-speech-3-3-8b"


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": "688e7a943167401c310f0975cb68f1a35e0bddc3b65f60bde89c37860e07edf1", "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 "".join(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")


print(create_and_wait_for_prediction({
    "audio": [
        "https://blog.api.market/wp-content/uploads/2026/10/granite-speech-3-3-8b-input-api-market.wav"
    ],
    "top_k": 50,
    "top_p": 0.9,
    "prompt": "Transcribe the speech into written form.",
    "max_tokens": 512,
    "min_tokens": 0,
    "temperature": 0.6,
    "presence_penalty": 0,
    "frequency_penalty": 0
}))
```

## FAQ
- **How do I tell the model what to do with the audio?**
  - Put the instruction in `prompt`, for example "Transcribe the speech into written form." The audio goes in `audio` as an array of URLs.
- **Why is the output a list?**
  - The model returns its answer as text chunks. Join them without a separator to get the full text.
- **What am I billed for?**
  - Output tokens only: $10 per million, in units of 100 tokens ($0.001), rounded up per request. Audio and prompt tokens are not billed.
- **How long can the answer be?**
  - Up to `max_tokens` tokens (default 512). Raise it for long recordings; the cost grows with the number of output tokens.
- **Is the 201 response from POST an error?**
  - No. 201 Created means the prediction was accepted. Use the returned `id` to poll GET `/predictions/{id}` until `status` is `succeeded`, `failed` or `canceled`.
- **What if my prediction fails?**
  - Failed predictions are not billed. Read the `error` field, check that the audio URL is reachable and try again.

## Contact Us
If you have any questions, reach out to us through the chat on the bottom right of the page. You can also email our support team at hello@magicapi.com.


Transcribe and translate speech with IBM Granite Speech 3.3 8B: send audio URLs and a prompt, get text back. Pay as you go, no monthly fee.