# Make a talking video with Speak

> Use the Nodespell Speak Creative Tool to make a portrait talk from a script or audio, or lip-sync an existing video to new speech.

Source: https://docs.nodespell.com/creative-tools/speak/ · Last updated: 2026-11-10

Speak is the Nodespell Creative Tool for talking videos. It covers two jobs: making a still portrait or character speak from a script or an audio file (avatar models), and re-syncing the mouth of a person in an existing video to new audio (lip-sync models). The selected model decides which inputs you need.

## When to use Speak

| You have…                                         | Choose a Speak model that takes…                         |
| ------------------------------------------------- | -------------------------------------------------------- |
| A portrait image and a script                     | An image and a script, such as P-Video Avatar            |
| A portrait image and recorded or generated speech | An image and audio, such as P-Video Avatar or Omni Human |
| A video of someone talking and new audio          | A video and audio, such as Sync Lipsync V3               |

If you only need the voice, without a picture, use [Voice Generator](https://docs.nodespell.com/creative-tools/voice-generator/). For a general shot from a still, use [Image to Video](https://docs.nodespell.com/creative-tools/image-to-video/).

## Make a portrait speak

1. Switch the workspace to **Create**. Select **Image to Video** in the tool rail, open the tool switcher at the top of the form, and choose **Speak**. It is also under **All tools**.
2. Keep the default model, **P-Video Avatar**, or choose another.
3. Add the portrait to **Source image**.
4. Either type the words in **Voice script**, or add an audio file to **Reference audio**. If you add audio, P-Video Avatar uses it instead of the generated voice.
5. For a script, choose a **Voice** and adjust the voice language and voice prompt if needed.
6. Set **Resolution** (720p or 1080p) and **Generations** from 1 to 4.
7. Select **Generate**.

## Lip-sync an existing video

1. Open **Speak** and choose a lip-sync model, such as **Sync Lipsync V3**.
2. Add the video of the person talking and the new audio track. Both are required on Sync Lipsync V3.
3. Choose how to handle a length mismatch with **Sync Mode**: Cut Off, Loop, Bounce, Silence or Remap.
4. Select **Generate**.

To create the new audio first, use [Voice Generator](https://docs.nodespell.com/creative-tools/voice-generator/) and pick its result in the audio field. See [Make a talking video](https://docs.nodespell.com/recipes/talking-video/) for the full chain.

## Speak controls

| Control                             | What it does                                        | Options and defaults      |
| ----------------------------------- | --------------------------------------------------- | ------------------------- |
| Tool switcher (title)               | Moves to another video Creative Tool                | All video tools           |
| Model picker                        | Chooses an avatar or lip-sync model                 | Default: P-Video Avatar   |
| **Previous model** / **Next model** | Steps through eligible models                       | —                         |
| **More settings**                   | Less-used model settings                            | Collapsed                 |
| **Generations**                     | Number of runs submitted together                   | 1 to 4, default 1         |
| **Generate**                        | Adds or updates the model node and submits the runs | Shows the credit estimate |

## P-Video Avatar settings

| Setting             | What it does                                                  | Default                  |
| ------------------- | ------------------------------------------------------------- | ------------------------ |
| **Source image**    | The portrait or character that speaks                         | Required                 |
| **Reference audio** | Speech to use instead of a generated voice                    | Optional                 |
| **Voice script**    | Words to say when no audio is provided                        | Empty                    |
| **Resolution**      | 720p or 1080p                                                 | 720p                     |
| **Voice**           | Generated voice preset                                        | Zephyr (Female)          |
| **Voice Language**  | Language or accent, such as English (US), Spanish or Japanese | English (US)             |
| **Voice Prompt**    | How to deliver the script; not spoken                         | "Say the following."     |
| **Video Prompt**    | How the person should look or behave while speaking           | "The person is talking." |

## Recommended models for Speak

The **Featured** collection lists P-Video Avatar (default), HeyGen Avatar V and Sync Lipsync V3. **All models** also includes Davinci Magihuman, Omni Human, Kling Lip Sync, Sync Lipsync 2, Sync Lipsync 2 Pro, HeyGen V3 Lipsync Precision and HeyGen V3 Lipsync Speed.

| Model             | Inputs                                                                  |
| ----------------- | ----------------------------------------------------------------------- |
| P-Video Avatar    | Image, plus script or audio                                             |
| HeyGen Avatar V   | Script, plus an existing avatar ID and voice ID from that service       |
| Sync Lipsync V3   | Video and audio                                                         |
| Omni Human        | Image and audio; its description recommends audio of 15 seconds or less |
| Davinci Magihuman | Image and prompt, with optional driving audio; 1–30 seconds             |

See [Choose a model](https://docs.nodespell.com/creative-tools/choose-a-model/) for filters and switching.

## What changes with the model?

- **Image or video source:** avatar models animate a still; lip-sync models change an existing clip.
- **Script or audio:** some models generate speech from text, others need an audio file.
- **Length:** usually follows the audio or script. Some models note limits, such as Kling Lip Sync's 2–10 second source videos.
- **Resolution:** set on some models (P-Video Avatar, HeyGen Avatar V); others follow the source.

## Where does the video go?

**Generate** places the Speak model node in the current graph level and runs it. The talking video appears in **Results** in the Create form and on the node; see [Takes](https://docs.nodespell.com/graph/takes/). Generated media also appears under **Generated** in the [Asset Manager](https://docs.nodespell.com/assets/asset-manager/).

## Common problems

- **The mouth does not match the words.** Use a clear, front-facing face with good lighting. Side views and covered mouths reduce quality.
- **"This value is required."** A required image, video or audio is missing for the selected model.
- **The voice sounds wrong.** Change **Voice**, **Voice Language** or **Voice Prompt**, or record or generate the audio separately and use it as **Reference audio**.

## Frequently asked questions

### Can Speak make a photo talk?

Yes. Use an avatar model such as P-Video Avatar, add the photo as the source image, and supply a script or an audio file.

### Can Speak change what someone says in an existing video?

Yes, with a lip-sync model such as Sync Lipsync V3. Add the video and the new audio, and the model re-times the mouth to match.

### Do I need to record audio for Speak?

No. P-Video Avatar can generate speech from **Voice script**. Lip-sync models need an audio file, which you can make in [Voice Generator](https://docs.nodespell.com/creative-tools/voice-generator/).

### What is the default Speak model?

The default is P-Video Avatar, which turns a portrait image plus a script or audio into a talking video at 720p or 1080p.

## Related

- [Make a talking video](https://docs.nodespell.com/recipes/talking-video/) — script to voice to speaking video
- [Voice Generator](https://docs.nodespell.com/creative-tools/voice-generator/) — create the speech track
- [Image to Video](https://docs.nodespell.com/creative-tools/image-to-video/) — animate a still without speech
- [Choose a model](https://docs.nodespell.com/creative-tools/choose-a-model/) — compare avatar and lip-sync models
- [Timeline](https://docs.nodespell.com/edit/timeline/) — place the talking clip in an edit