---
title: "AI audio models"
description: "Every AI audio generation model on Hedra, grouped by the lab that makes it."
canonical: "https://www.hedra.com/models/audio"
---

# AI audio models

[View on Hedra](https://www.hedra.com/models/audio)

## [ElevenLabs Audio Isolation](https://www.hedra.com/models/audio/elevenlabs/audio-isolation.md)

Strip background noise from a recording, keeping the speech.

12¢/1 minute; a recording shorter than 4.65 seconds bills 4.65 seconds.

## [ElevenLabs English STS V2](https://www.hedra.com/models/audio/elevenlabs/english-sts-v2.md)

16.8¢/1 minute.

## [ElevenLabs Flash Multilingual V2](https://www.hedra.com/models/audio/elevenlabs/flash-multilingual-v2.md)

The low-latency voice across 30+ languages, for interactive and high-volume speech.

10¢/2,000 characters; plus up to 25¢/1,000 characters for a Voice Library voice that sets its own rate.

## [ElevenLabs Flash V2](https://www.hedra.com/models/audio/elevenlabs/flash-v2.md)

The low-latency English voice, for interactive speech.

10¢/2,000 characters; plus up to 25¢/1,000 characters for a Voice Library voice that sets its own rate.

## [ElevenLabs Multilingual STS V2](https://www.hedra.com/models/audio/elevenlabs/multilingual-sts-v2.md)

16.8¢/1 minute.

## [ElevenLabs Multilingual V2](https://www.hedra.com/models/audio/elevenlabs/multilingual-v2.md)

Steady, natural narration across 30+ languages, for finished voiceover.

10¢/1,000 characters; plus up to 25¢/1,000 characters for a Voice Library voice that sets its own rate.

## [ElevenLabs Music](https://www.hedra.com/models/audio/elevenlabs/music.md)

Full tracks from a written brief, with optional lyrics placed across the length you ask for.

60¢/minute; a partial minute bills whole.

## [ElevenLabs Sound Effects](https://www.hedra.com/models/audio/elevenlabs/sound-effects.md)

One-off sound effects from a written description, loopable on request.

0.71¢/second.

## [ElevenLabs V3](https://www.hedra.com/models/audio/elevenlabs/v3.md)

The most expressive ElevenLabs voice — emotional range and delivery cues for performance, not just narration.

10¢/1,000 characters; plus up to 25¢/1,000 characters for a Voice Library voice that sets its own rate.

## [ElevenLabs Voice Clone](https://www.hedra.com/models/audio/elevenlabs/voice-clone.md)

Use an audio clip to create a new Voice.

Free.

## [MiniMax Speech-2.5 HD Preview](https://www.hedra.com/models/audio/minimax/speech-25-hd-preview.md)

The high-fidelity tier — closest voice likeness, across 40+ languages.

12.86¢/1,000 characters; each Chinese character counts as 2 characters.

## [MiniMax Speech-2.5 Turbo Preview](https://www.hedra.com/models/audio/minimax/speech-25-turbo-preview.md)

The value tier — natural English delivery across 40+ languages, at a lower rate.

7.14¢/1,000 characters; each Chinese character counts as 2 characters.

[Model API documentation](https://www.hedra.com/docs/pages/developer/v3/model-jobs.md)