# Script → narration

Turn a short script into a separate voiceover track, with a model and voice you choose.

Source: https://www.creativly.ai/docs/flow/text-to-audio

![Actual dark Flow workspace: an Aurora narration script connected to a completed Gemini 2.5 Flash TTS audio node, with Kore selected and an eight-second audio player.](https://pub-60e30bed45514f03933d3900731bea31.r2.dev/public-app-sessions/v1/documentation/flow-audio-workspace-88c82b8c919e225b.webp)

A Text script feeds an Audio node. The model, voice, connection and finished player are all visible.

Make narration for a product film, a concept presentation, or a short social clip. This creates **an audio file**; it does not automatically add a soundtrack to a video.

## Connect your script

1. Right-click empty canvas and choose **Prompt**. Write the words you want spoken in the new Text node.
2. Add an **Audio** node. Drag the Text node's right-hand output to Audio's left-hand **Text** input.
3. Select Audio, choose a model and voice, then press **Run** below the node.

Our script:

> Meet Aurora. A quiet moment, a warmer morning. Crafted for the small rituals that make every day feel considered.

**One script is enough.** Here, Audio's own prompt box stays empty because Text supplies the words. You can also write directly in Audio without a Text node. If you fill both, Flow combines them.

## Choose the voice

![Gemini 2.5 Flash TTS model control beside its open voice menu, with Kore selected and other voices listed.](https://pub-60e30bed45514f03933d3900731bea31.r2.dev/public-app-sessions/v1/documentation/flow-audio-voice-c78ca3f9f36d8126.webp)

This example uses Gemini 2.5 Flash TTS with Kore. Voice choices depend on the model.

Check **Run nodes with** and the current credit estimate before generating. Changing a voice does not change an existing recording—run the Audio node again when you want a new take.

## Listen to the result

The actual Gemini output from this workflow:

[Listen to Aurora narration, generated with Gemini 2.5 Flash TTS and Kore](https://pub-60e30bed45514f03933d3900731bea31.r2.dev/public-app-sessions/v1/documentation/flow-narration-preview-b88f2a9a60108334.m4a)

[Download the original WAV](https://pub-60e30bed45514f03933d3900731bea31.r2.dev/public-app-sessions/v1/documentation/flow-narration-original-7d141f863866ad29.wav). The inline player uses a smaller listening copy; the download is unchanged.

On the canvas, use **Play**, the seek bar, volume, and the download icon to review and save a take.

### If a downloaded clip has the wrong extension

A current export issue can give a restored WAV result an `.mp3` filename. It does not convert the audio. The original WAV linked above is correctly named; check exported files before handing them off.

[Build a product film](https://www.creativly.ai/docs/flow/image-to-video)

Create the visual clip separately, then bring video and narration into an editor.

Open in Creativly: [Flow ↗](https://www.creativly.ai/app/flow)