Understanding voice models

If you’ve ever wanted your content to sound more natural, more personal, or more closely aligned with a particular brand voice, voice models make that possible.

In HeyGen, voices bring your scripts and avatars to life with AI-generated narration. Each voice you select is powered by a voice engine: a model designed to produce natural, expressive speech that accurately reflects the chosen language. Selecting the right engine helps you match the tone, pace and emotion of your message.

What voice models are

A voice model is the underlying AI system that generates speech. It determines how a voice sounds, how expressive it is, how quickly it speaks, and how well it performs across languages.

HeyGen offers multiple voice engines, each optimised for different uses such as training, marketing, storytelling or localisation.

Automatic voice engine

The Auto setting allows HeyGen to select the most suitable voice engine automatically based on your video’s language and content. This is a good option if you want reliable results without having to choose a model manually.

ElevenLabs voice engine

ElevenLabs provides studio-quality narration in more than 70 languages, making it well suited to most video and voice projects.

If you’re using a bespoke voice, you can also choose which voice model powers it, giving you greater control over tone and realism. By default, HeyGen uses ElevenLabs’ multilingual V3 model, which is known for natural expression and strong performance across languages.

Turbo voice models

For projects that require faster generation, you can switch to one of the Turbo models. These provide lower latency and quicker processing, but are primarily optimised for English-language content.

Starfish voice engine

Starfish is optimised for Asian languages, including Chinese, Japanese and Korean. It ensures natural pronunciation and pacing for content tailored to each region.

Panda voice engine

Panda is HeyGen’s expressive engine, designed for emotive delivery and advanced control. It supports features such as Voice Director and Voice Mirroring, providing precise control over timing, emphasis and tone.

Fish voice engine

Fish, powered by fish.audio, focuses on expressive English voiceovers. It works well for storytelling, conversational videos and content that benefits from nuanced delivery.

Together, voice engines and models give you control over how your videos sound, from tone and emotion to pace and linguistic accuracy.