Name: Dia 1.6B API: Realistic Dialogue Text to Speech
Brand: magicapi
SKU: api-magicapi-realistic-dialogue-audio-generation-zsxkib-dia
Availability: InStock
Author: API Market

Shashank Agarwal

# 🔊 Dialogue Text-to-Speech API (Nari Labs Dia 1.6B)

## Overview
Dia 1.6B is a dialogue text-to-speech model from Nari Labs. Its developer describes it as generating realistic dialogue audio from text, including non-verbal cues and voice cloning.

Write a script with speaker tags (`[S1]`, `[S2]`) and non-verbal cues in parentheses, such as `(laughs)` or `(whispers)`. The model returns the URL of one WAV file with the whole conversation. Add an `audio_prompt` to make the voices follow a sample recording.

## Features
- **Multi-speaker dialogue**: mark each line with `[S1]` or `[S2]` to switch speakers in one request.
- **Non-verbal cues**: write cues in parentheses, for example `(laughs)` or `(whispers)`.
- **Voice cloning**: pass a `.wav`, `.mp3` or `.flac` sample as `audio_prompt` (and its transcript as `audio_prompt_text`) and the model will attempt to mimic that voice style. Samples are cut to `max_audio_prompt_seconds` (1 to 120, default 10).
- **Length control**: `max_new_tokens` (500 to 4096, default 3072) sets the maximum audio length; 86 tokens is about 1 second, so 4096 tokens is about 48 seconds.
- **Speed control**: `speed_factor` (0.5 to 1.5, default 1) slows down or speeds up playback.
- **Generation controls**: `cfg_scale` (1 to 5, default 3), `temperature` (1 to 2.5, default 1.8), `top_p` (0.1 to 1, default 0.95), `cfg_filter_top_k` (10 to 100, default 45) and `seed`.

## Use Cases
1. **Podcast and interview drafts** with two voices from one script.
2. **Dialogue for games and animation** with laughs, whispers and other reactions.
3. **Explainer and ad voice-overs** written as a conversation.
4. **Consistent voices**: reuse one `audio_prompt` sample across many scripts.

## Visual Examples
### Example Input
```json
{
  "text": "[S1] Have you tried the new garden planner yet? [S2] Not yet. Is it any good? [S1] It's great. You sketch the beds and it tells you what to plant and when. (laughs) [S2] Okay, I'm sold.",
  "top_p": 0.95,
  "cfg_scale": 4,
  "temperature": 1.3,
  "speed_factor": 0.94,
  "max_new_tokens": 3072,
  "cfg_filter_top_k": 35
}
```

### Example Output
[Listen to the example output](https://blog.api.market/wp-content/uploads/2026/10/dia-output-ayrfvt4b-api-market.mp3)

Each request returns one audio file URL in `output`, with all speakers and cues in a single track.

## Input Parameters
| Field | Type | Required | Default | Allowed values | Description |
| --- | --- | --- | --- | --- | --- |
| `text` | string | Yes |  |  | Input text for dialogue generation. Use `[S1]`, `[S2]` to indicate different speakers and a description in parentheses for non-verbal cues, e.g. `(laughs)`, `(whispers)` |
| `audio_prompt` | string (URL) | No | none |  | Optional audio file (.wav/.mp3/.flac) for voice cloning. The model will attempt to mimic this voice style |
| `audio_prompt_text` | string | No | none |  | Optional transcript of the audio prompt. If provided, it is prepended to the main text input |
| `max_new_tokens` | integer | No | `3072` | 500 to 4096 | Controls the length of generated audio. Higher values create longer audio (86 tokens ≈ 1 second of audio) |
| `max_audio_prompt_seconds` | integer | No | `10` | 1 to 120 | Maximum duration in seconds of the voice cloning audio prompt. Only used when an audio prompt is provided. Longer samples are truncated to this length |
| `cfg_scale` | number | No | `3` | 1 to 5 | Controls how closely the audio follows your text. Higher values (3 to 5) follow the text more strictly; lower values may sound more natural but deviate more |
| `temperature` | number | No | `1.8` | 1 to 2.5 | Controls randomness in generation. Higher values (1.3 to 2.0) increase variety; lower values make output more consistent |
| `top_p` | number | No | `0.95` | 0.1 to 1 | Controls diversity of word choice. Higher values include more unusual options. Most users do not need to adjust this |
| `cfg_filter_top_k` | integer | No | `45` | 10 to 100 | Technical parameter for filtering audio generation tokens. Higher values allow more diverse sounds; lower values create more consistent audio |
| `speed_factor` | number | No | `1` | 0.5 to 1.5 | Adjusts playback speed of the generated audio. Values below 1.0 slow it down; 1.0 is original speed |
| `seed` | integer | No | random |  | Random seed for reproducible results. Use the same seed to get the same output for identical inputs |

## Pricing
- **Pay As You Go (PAYG) pricing**: you only pay for what you use.
- **Price**: $0.038 per API unit.
- **1 API unit = 1 successful request with up to 35 seconds of processing.** A longer job bills one more unit for each started block of 35 seconds (for example, 40 seconds of processing is 2 units, $0.076).
- Processing time grows with the length of `text` and `max_new_tokens`.
- **Example cost**: the example request (a four-line, two-speaker script, `max_new_tokens` 3072) took 14.3 seconds of processing, which is 1 unit, so $0.038.
- **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/realistic-dialogue-audio-generation-zsxkib-dia`

### 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/realistic-dialogue-audio-generation-zsxkib-dia/predictions" \
  -H "Content-Type: application/json" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY" \
  -d '{
    "version": "2119e338ca5c0dacd3def83158d6c80d431f2ac1024146d8cca9220b74385599",
    "input": {
      "text": "[S1] Have you tried the new garden planner yet? [S2] Not yet. Is it any good? [S1] It'\''s great. You sketch the beds and it tells you what to plant and when. (laughs) [S2] Okay, I'\''m sold.",
      "top_p": 0.95,
      "cfg_scale": 4,
      "temperature": 1.3,
      "speed_factor": 0.94,
      "max_new_tokens": 3072,
      "cfg_filter_top_k": 35
    }
  }'
```

Response (201 Created):
```json
{
  "id": "ayrfvt4bbxrmt0d1d9jvb700cm",
  "version": "2119e338ca5c0dacd3def83158d6c80d431f2ac1024146d8cca9220b74385599",
  "status": "starting",
  "created_at": "2026-10-23T20:13:22.294Z"
}
```

#### Check Prediction Status
```bash
curl -X GET "https://prod.api.market/api/v1/magicapi/realistic-dialogue-audio-generation-zsxkib-dia/predictions/{id}" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY"
```

Response when the prediction has finished (200 OK):
```json
{
  "id": "ayrfvt4bbxrmt0d1d9jvb700cm",
  "status": "succeeded",
  "output": "https://blog.api.market/wp-content/uploads/2026/10/dia-output-ayrfvt4b-api-market.mp3",
  "error": null,
  "created_at": "2026-10-23T20:13:22.294Z",
  "started_at": "2026-10-23T20:14:27.506Z",
  "completed_at": "2026-10-23T20:14:41.821Z"
}
```

`output` is a single URL string pointing to the generated audio file.

### 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 missing `text` or a `max_new_tokens` outside 500 to 4096.
- **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/realistic-dialogue-audio-generation-zsxkib-dia";

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: "2119e338ca5c0dacd3def83158d6c80d431f2ac1024146d8cca9220b74385599", 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({
  "text": "[S1] Have you tried the new garden planner yet? [S2] Not yet. Is it any good? [S1] It's great. You sketch the beds and it tells you what to plant and when. (laughs) [S2] Okay, I'm sold.",
  "top_p": 0.95,
  "cfg_scale": 4,
  "temperature": 1.3,
  "speed_factor": 0.94,
  "max_new_tokens": 3072,
  "cfg_filter_top_k": 35
}).then((output) => console.log(output));
```

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

API_KEY = "YOUR_API_MARKET_KEY"
BASE_URL = "https://prod.api.market/api/v1/magicapi/realistic-dialogue-audio-generation-zsxkib-dia"


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": "2119e338ca5c0dacd3def83158d6c80d431f2ac1024146d8cca9220b74385599", "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")


print(create_and_wait_for_prediction({
    "text": "[S1] Have you tried the new garden planner yet? [S2] Not yet. Is it any good? [S1] It's great. You sketch the beds and it tells you what to plant and when. (laughs) [S2] Okay, I'm sold.",
    "top_p": 0.95,
    "cfg_scale": 4,
    "temperature": 1.3,
    "speed_factor": 0.94,
    "max_new_tokens": 3072,
    "cfg_filter_top_k": 35
}))
```

## FAQ
- **How do I write a conversation?**
  - Start each line with a speaker tag, `[S1]` or `[S2]`, and put non-verbal cues in parentheses where they happen, for example `(laughs)`.
- **How do I clone a voice?**
  - Pass a short `.wav`, `.mp3` or `.flac` sample as `audio_prompt` and its transcript as `audio_prompt_text`. Only the first `max_audio_prompt_seconds` (default 10) of the sample is used. Only use recordings of people who have agreed to it.
- **How long can the audio be?**
  - `max_new_tokens` caps the length: 86 tokens is about 1 second, so the maximum of 4096 is about 48 seconds. Split longer scripts into several requests.
- **Why does the same script sound different each time?**
  - Generation is random by default. Set `seed` to repeat a result, or lower `temperature` for more consistent audio.
- **What if my prediction fails?**
  - Failed predictions are not billed. Read the `error` field, adjust the input 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.


Generate dialogue audio with Nari Labs Dia 1.6B: [S1]/[S2] speaker tags, non-verbal cues such as (laughs), and voice cloning from a sample.