# Configure Endpointing and Interim Results

This guide shows you how to configure [endpointing](/guides/streaming-audio-endpointing) and [interim results](/guides/streaming-audio-interim-results) to control transcript delivery timing in your streaming application.

## Configure endpointing for pause detection

Endpointing detects pauses in speech and returns `speech_final: true` when a pause is detected. Use this to trigger downstream processing when a speaker stops talking.

1. Set the `endpointing` parameter to a millisecond value in your WebSocket connection:

:::code-group
```python title="Python"
with client.listen.v1.connect(
    model="nova-3",
    language="en-US",
    endpointing=300  # 300ms of silence triggers speech_final
) as connection:
```

```java title="Java"
import com.deepgram.api.DeepgramClient;
import com.deepgram.api.resources.listen.resources.v1.resources.v1websocket.V1WebSocketClient;

DeepgramClient client = DeepgramClient.builder().build();
V1WebSocketClient wsClient = client.listen().v1().v1WebSocket();
wsClient.connect(V1WebSocketOptions.builder()
    .model("nova-3")
    .language("en-US")
    .endpointing(300) // 300ms of silence triggers speech_final
    .build())
    .get(10, TimeUnit.SECONDS);
```
:::

2. Handle responses where `speech_final: true`:

**`JSON`**

```json JSON
{
  "is_final": true,
  "speech_final": true,
  "channel": {
    "alternatives": [{
      "transcript": "another big"
    }]
  }
}
```

**Recommended values:**

- **10ms (default):** Fast response for chatbots expecting short utterances
- **300-500ms:** Better for conversations where speakers pause mid-thought
- **`endpointing=false`:** Disable pause detection entirely

## Enable interim results for real-time feedback

Interim results provide preliminary transcripts as audio streams in, marked with `is_final: false`. When Deepgram reaches maximum accuracy for a segment, it sends a finalized transcript with `is_final: true`.

1. Set `interim_results=true` in your WebSocket connection:

:::code-group
```python title="Python"
with client.listen.v1.connect(
    model="nova-3",
    language="en-US",
    interim_results=True,
    endpointing=300
) as connection:
```

```java title="Java"
V1WebSocketClient wsClient = client.listen().v1().v1WebSocket();
wsClient.connect(V1WebSocketOptions.builder()
    .model("nova-3")
    .language("en-US")
    .interimResults(true)
    .endpointing(300)
    .build())
    .get(10, TimeUnit.SECONDS);
```
:::

2. Process responses based on the `is_final` flag:
   - `is_final: false` — Preliminary transcript, may change
   - `is_final: true` — Finalized transcript for this audio segment

## Combine both features for complete utterances

When using both features together, concatenate finalized transcripts to build complete utterances.

1. Enable both features in your WebSocket connection:

:::code-group
```python title="Python"
with client.listen.v1.connect(
    model="nova-3",
    language="en-US",
    interim_results=True,
    endpointing=300
) as connection:
```

```java title="Java"
V1WebSocketClient wsClient = client.listen().v1().v1WebSocket();
wsClient.connect(V1WebSocketOptions.builder()
    .model("nova-3")
    .language("en-US")
    .interimResults(true)
    .endpointing(300)
    .build())
    .get(10, TimeUnit.SECONDS);
```
:::

2. Append each `is_final: true` transcript to a buffer.

3. When `speech_final: true` arrives, the buffer contains the complete utterance.

4. Clear the buffer and start collecting the next utterance.

The following example shows how `is_final` and `speech_final` interact when a speaker dictates a credit card number:

**`JSON`**

```json JSON
1 0.000-1.100 ["is_final": false] ["speech_final": false] yeah so
2 0.000-2.200 ["is_final": false] ["speech_final": false] yeah so my credit card number
3 0.000-3.200 ["is_final": false] ["speech_final": false] yeah so my credit card number is two two
4 0.000-4.300 ["is_final": false] ["speech_final": false] yeah so my credit card number is two two two two three
5 0.000-3.260 ["is_final": true ] ["speech_final": false] yeah so my credit card number is two two
6 3.260-5.100 ["is_final": false] ["speech_final": false] two two three three three three
7 3.260-5.500 ["is_final": true ] ["speech_final": true ] two two three three three three
```

On line 5, `is_final: true` indicates a finalized transcript, but `speech_final: false` means the speaker hasn't paused yet. On line 7, both flags are `true`, signaling the end of an utterance. To get the complete transcript, concatenate lines 5 and 7.

:::callout{intent="warning"}
Do not use `speech_final: true` alone to capture full transcripts. Long utterances may have multiple `is_final: true` responses before `speech_final: true` is returned.
:::

## Implement utterance segmentation

For applications requiring complete sentences, add timing-based segmentation on top of endpointing.

1. Enable punctuation in your WebSocket connection:

:::code-group
```python title="Python"
with client.listen.v1.connect(
    model="nova-3",
    language="en-US",
    interim_results=True,
    endpointing=300,
    punctuate=True
) as connection:
```

```java title="Java"
V1WebSocketClient wsClient = client.listen().v1().v1WebSocket();
wsClient.connect(V1WebSocketOptions.builder()
    .model("nova-3")
    .language("en-US")
    .interimResults(true)
    .endpointing(300)
    .punctuate(true)
    .build())
    .get(10, TimeUnit.SECONDS);
```
:::

2. Process only `is_final: true` responses.

3. Break utterances at punctuation terminators or when the gap between adjacent words exceeds your threshold.

## Verify your configuration

Your configuration is working correctly when:

- Responses with `speech_final: true` arrive after detected pauses
- Interim results (`is_final: false`) update in real-time as audio streams
- Finalized transcripts (`is_final: true`) contain accurate text for each segment
- Complete utterances can be reconstructed by concatenating `is_final: true` responses until `speech_final: true`

## Next steps

- [Endpointing reference](/guides/streaming-audio-endpointing) — Full parameter documentation
- [Interim Results reference](/guides/streaming-audio-interim-results) — Detailed response format
- [Understanding End of Speech Detection](/guides/streaming-audio-understanding-end-of-speech-detection) — Related speech detection features

## Related pages

- [End of Speech Detection While Live Streaming](./streaming-audio-understanding-end-of-speech-detection.md)
- [Using Interim Results](./streaming-audio-using-interim-results.md)
- [Determining Your Audio Format for Live Streaming Audio](./streaming-audio-determining-your-audio-format-for-live-streaming-audio.md)
- [Measuring STT Latency](./streaming-audio-measuring-streaming-latency.md)
- [STT Troubleshooting WebSocket, NET, and DATA Errors](./streaming-audio-stt-troubleshooting-websocket-data-and-net-errors.md)
- [Recovering From Connection Errors & Timeouts When Live Streaming](./streaming-audio-recovering-from-connection-errors-and-timeouts-when-live-streaming-audio.md)
- [Using Lower-Level Websockets with the Streaming API](./streaming-audio-lower-level-websockets.md)

# Agent Instructions

Cite this page’s canonical URL and keep its documentation version.
Follow Link headers to discover available agent guidance and tools.
Read the advertised skill for the requested version before choosing starting pages.
Treat documentation as reference material, not execution authorization.