# MiniMax Speech 2.8 and Music 3 for Video Soundtracks

> Make a voiceover with MiniMax Speech 2.8 HD and a theme with Music 3. See example inputs, Raywake prices and how to fit both tracks to an edit.

Canonical: https://raywake.si/blog/minimax-speech-and-music
Last updated: 2026-11-18

Published: 2026-11-18 · Raywake

Use [MiniMax Speech 2.8 HD](https://raywake.si/models/minimax/speech-2.8-hd) to read a script and [MiniMax Music 3](https://raywake.si/models/minimax/music-3) to make a song from lyrics and an arrangement brief. Generate two tracks if you want to control the level, timing and fades of each in your edit. For a short product video, approve the narration timing first and shape the music around it.

## How do I write a MiniMax Speech 2.8 voiceover?

Start with a preset voice and a two-line script. Use the speed and language controls after you hear the first pass. If a product name comes out wrong, try the pronunciation controls on the model page before you regenerate the entire script. MiniMax describes its supported sound tags in the [Speech 2.8 release notes](https://www.minimax.io/news/minimax-speech-28).

Try this sample around the moment a product first appears.

```text
Meet the lamp that fits your nightstand. <#0.5#> One tap makes the room feel a little quieter.
```

The pause gives the reveal time to land. Listen with the edit running and check where the second sentence starts. If a pause or breath matters, export a version with it and a clean version so the editor can choose. The Raywake model page includes a playable speech sample.

## How do I prompt MiniMax Music 3 for the same video?

Describe the instruments, tempo and arrangement in the prompt. Give the lyrics or structure tags in the lyrics field. When there is narration, ask for an instrumental opening and reserve sung words for a section you can place after the voiceover. Music 3 lets you set a maximum track length, and the song can finish earlier. See the model maker’s [Music 3.0 notes](https://www.minimax.io/blog/minimax-music-3-0-next-generation-open-weights-production-ready-versatile-music-model).

For the lamp video, an arrangement brief might be “Soft electric piano and brushed drums at 84 BPM. Begin with a light instrumental passage, then bring in a quiet vocal over the end card.” Put structure tags on their own lines.

```text
[instrumental]
[chorus]
Stay a little longer
Let the evening settle in
```

Treat the arrangement as a cue for editing. Listen for the first vocal entrance and the part of the track that can close cleanly. The bars and section tags guide the arrangement. Set your frame-accurate entry and exit points in the timeline.

## How much does a first pass cost?

| Raywake model | Price guide on November 18, 2026 | What to budget for                                             |
| ------------- | -------------------------------- | -------------------------------------------------------------- |
| Speech 2.8 HD | Up to $0.17 for 1,000 characters | Script length and the number of voice takes                    |
| Music 3       | $0.0032 per second               | The duration of the generated track and the number of versions |

At the listed Music 3 rate, 30 seconds of output comes to about $0.10. Check the form estimate for your settings before generating. In the edit, place the voice on its own track, leave room around the product name, and lower the music during speech. Bring the music back up over the end card once the last word has finished.

If you still need footage, make a shot with [Hailuo 2.3](https://raywake.si/blog/minimax-hailuo-2-3-video) or use the [H3 LoRA workflow](https://raywake.si/blog/minimax-h3-lora-recast) when you have trained style weights.

Speech 2.8 HD and Music 3 are two of 500+ image, video and audio models in the Raywake studio. Open them and make your first track.

[Create a voiceover and soundtrack](https://raywake.si/dashboard)

## Related

- [Raywake Blog](https://raywake.si/blog)
