> For the complete documentation index, see [llms.txt](https://docs.convai.com/api-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.convai.com/api-docs/plugins-and-integrations/convai-unreal-engine-plugin/features/lip-sync/record-and-replay-lip-sync.md).

# Record and replay lip sync

Capture a live lip-sync sequence during a Convai conversation and replay it later in cutscenes, cached responses, or offline previews without re-streaming.

The `UConvaiFaceSyncComponent` recording API lets you capture the blendshape frames that arrive during a live conversation and replay them on demand without re-streaming from Convai. Use it for cutscenes where the same lip-sync sequence plays back repeatedly, for pre-warming a response before it is triggered, or for debugging and offline preview of facial animation.

{% hint style="warning" %}
The recording API — `StartRecordingLipSync`, `FinishRecordingLipSync`, `PlayRecordedLipSync`, `IsPlaying`, and `GetCurrentFrame` — is C++ only. These methods have no `UFUNCTION` decoration and are not accessible from Blueprints.
{% endhint %}

### When to use the recording API

| Use case                                | Approach                                                                            |
| --------------------------------------- | ----------------------------------------------------------------------------------- |
| Cutscene with a fixed character speech  | Record once during development, store the sequence, replay in the cutscene          |
| Pre-warmed response cache               | Record the response on first play, replay for subsequent identical responses        |
| Offline facial animation preview        | Record during a PIE session, inspect frame data with `GetCurrentFrame()`            |
| Live streaming (real-time conversation) | Do not use the recording API — the AnimGraph node handles live frames automatically |

### Prerequisites

* The Actor has a `Convai Face Sync` component and a working `Convai Chatbot` component.
* The character already produces lip sync during conversation (see [Lip sync quick start](/api-docs/plugins-and-integrations/convai-unreal-engine-plugin/features/lip-sync/lip-sync-quick-start.md) if not).

### Record a sequence

{% stepper %}
{% step %}

#### Start recording before the speech turn

Call `StartRecordingLipSync()` on the `UConvaiFaceSyncComponent` instance. From this point, every blendshape frame delivered by Convai is buffered internally in addition to being applied to the face.

```cpp
// pseudocode
UConvaiFaceSyncComponent* FaceSync = MyCharacter->FindComponentByClass<UConvaiFaceSyncComponent>();
if (FaceSync)
{
    FaceSync->StartRecordingLipSync();
}
```

{% endstep %}

{% step %}

#### Let the character speak

Trigger a conversation as normal. The character speaks and its face animates. The component captures the incoming frames in the background — no additional calls are required while recording.
{% endstep %}

{% step %}

#### Stop recording and store the result

When the speech turn ends, call `FinishRecordingLipSync()`. The method stops capturing and returns an `FAnimationSequenceBP` containing the full list of frame-indexed blendshape frames, the total duration, and the frame rate.

```cpp
// pseudocode
FAnimationSequenceBP CachedSequence = FaceSync->FinishRecordingLipSync();
```

Store this value as a member variable on your Actor or component so you can replay it later. At this point you have the complete facial animation that played during the conversation turn.
{% endstep %}
{% endstepper %}

### Replay a recorded sequence

Call `PlayRecordedLipSync` on the `UConvaiFaceSyncComponent` instance to replay the stored sequence. The component feeds the recorded frames to the AnimGraph node exactly as it would during a live stream, with no active Convai conversation stream required.

#### PlayRecordedLipSync parameters

| Parameter           | Type                   | Description                                                                                           |
| ------------------- | ---------------------- | ----------------------------------------------------------------------------------------------------- |
| `RecordedLipSync`   | `FAnimationSequenceBP` | The sequence returned by `FinishRecordingLipSync`.                                                    |
| `StartFrame`        | `int32`                | Index of the first frame to replay. Pass `0` to start from the beginning.                             |
| `EndFrame`          | `int32`                | Index of the last frame to replay. Pass `-1` to play to the end of the sequence.                      |
| `OverwriteDuration` | `float`                | Override the sequence's stored playback duration in seconds. Pass `0.0` to use the original duration. |

`PlayRecordedLipSync` returns `bool` — `true` if the sequence started successfully. It returns `false` if the sequence is empty or invalid, recording is still active, `StartFrame` is greater than the last frame index, or `EndFrame` is greater than `0` and less than `StartFrame`.

```cpp
// pseudocode
bool bStarted = FaceSync->PlayRecordedLipSync(CachedSequence, 0, -1, 0.0f);
```

### Monitor playback state

Use these C++ methods to observe the component while a recorded sequence is playing:

| Method              | Returns              | Description                                                                                                                       |
| ------------------- | -------------------- | --------------------------------------------------------------------------------------------------------------------------------- |
| `IsPlaying() const` | `bool`               | `true` when the component is actively replaying frames from a recorded or live sequence.                                          |
| `GetCurrentFrame()` | `TMap<FName, float>` | The blendshape name-to-weight map for the frame currently being applied. Useful for driving secondary systems from the same data. |

### Example C++ flow

A typical cutscene usage looks like this:

```cpp
// pseudocode
// 1. Before or at the start of the speech turn:
FaceSync->StartRecordingLipSync();

// 2. The character speaks — recording happens automatically in the background.

// 3. When the speech turn ends:
FAnimationSequenceBP CachedSequence = FaceSync->FinishRecordingLipSync();

// 4. When you want to replay the sequence (e.g. on cutscene trigger):
FaceSync->PlayRecordedLipSync(CachedSequence, 0, -1, 0.0f);
```

{% hint style="info" %}
Replaying a sequence does not re-trigger audio. Play the matching audio clip separately and align playback start times to keep the face and voice in sync.
{% endhint %}

### Next steps

{% content-ref url="/pages/lsZCsc7Ypsg79ozXB0fr" %}
[Face Sync component reference](/api-docs/plugins-and-integrations/convai-unreal-engine-plugin/features/lip-sync/face-sync-component-reference.md)
{% endcontent-ref %}

{% content-ref url="/pages/n5M1eapXfiR0tOlawpbm" %}
[Lip sync usage examples](/api-docs/plugins-and-integrations/convai-unreal-engine-plugin/features/lip-sync/lip-sync-usage-examples.md)
{% endcontent-ref %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.convai.com/api-docs/plugins-and-integrations/convai-unreal-engine-plugin/features/lip-sync/record-and-replay-lip-sync.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.