How to Clone Your Voice in HeyGen (and Make It Sound Right)

A cloned voice makes an avatar video sound like you, or like your brand's narrator, in any script. HeyGen offers two levels of cloning, several voice engines behind them, and tools that fix the most common complaint: a clone that sounds like you but reads like a robot. This guide covers making a clone, recording it properly, improving it and directing delivery, following HeyGen's own help articles.

Instant or professional

HeyGen's in-house voice model, HeyGen Voice, makes voices in two modes:

The table follows HeyGen's developer documentation; the web app offers the same choice as Instant Voice Cloning or a professional clone. Each plan has a limit on the number of clones, and HeyGen notes you cannot buy extra slots in the app: delete voices you no longer need to free one.

Making an instant clone

1. In AI Studio, open the Voice section and click + New Voice, then Create New Voice.
2. Choose Instant Voice Cloning, then record or upload your sample.
3. Name the voice and save it to your Voice Library.
4. Optionally set it as the primary voice for your avatar (avatar group, three-dot menu, Set Primary Voice), so it is used by default.

You can also choose Generate a New Voice to design a synthetic voice from a description. HeyGen suggests starting with age, accent and gender, then adding detail; short, evocative prompts work best for character voices.

Recording a sample that clones well

HeyGen says clone quality depends directly on recording quality, and gives specific settings:

• iPhone: Settings, Voice Memos, Audio Quality, turn on Lossless Recording (off by default).
• Android: in the Voice Recorder app's settings, choose High Recording Quality (256 kbps, 48 kHz).
• External microphone: 6 to 8 inches from your mouth, not touching clothing; fabric noise is a common hidden cause of poor clones.
• At least 30 seconds of speech, in a quiet room, with minimal background noise.
• Speak, do not read. Aim for a natural conversational voice, using phrases that show your accent or dialect.

Improving a clone that sounds wrong

Improve Voice lets you re-record or upload a better sample, and compare the voice engines HeyGen supports for your voice. Open it from the Voice Library (three-dot menu next to the voice) or from AI Studio's Voice tab. Then:

1. Use the same test sentence for every engine, written like the content you normally make.
2. Listen for pronunciation, pacing, emotion, accent and similarity to your real voice.
3. Pick the most natural, consistent engine and save the configuration.

For a monotone clone, HeyGen's tip is to record a short, high-energy sample and add it as a Custom Emotion, which gives the model an expressive reference. If an accent is not preserved, HeyGen suggests a professional clone through ElevenLabs connected to HeyGen, choosing multilingual v2 for non-English content and turbo v2 for English.

Directing delivery: Voice Mirror and Direct Voice

• Voice Mirror copies your tone, pacing and emotion from a recording onto the chosen voice. In a scene, click Audio, record or upload, enable Voice Mirror and add it to the scene; HeyGen transcribes it for you. For an existing text scene, use Convert to Voice Mirroring and record against the teleprompter.
• Direct Voice shapes delivery without recording. Open the megaphone icon in the Editing Studio, pick a tone (Excited, Casual, Calm, Cool, Serious, Funny, Angry or Sarcastic) and edit its instruction, such as "speak in a calm, steady tone with a warm vibe". It works with custom clones and with public ElevenLabs and OpenAI voices.

Script formatting matters as much as the voice; see HeyGen script tips.

Choosing a stock voice instead

A clone is not always the right choice. For a brand narrator, a character or a language you do not speak, HeyGen's Voice Library of public voices may serve better. Filter by region, accent, emotion, age and use case, preview a few with your actual script, and mark favourites so the team uses the same voices across projects. If your script is in another language, HeyGen's Text Editor can translate it and you can pick a native voice for that language, which usually sounds more natural than your own clone speaking a language you do not.

Consent and consistency

Clone only your own voice or a voice whose owner has agreed in writing, and keep that consent on file with the project. For client work, record which voice, engine and settings each video used; a pick-up line months later only matches if the configuration is the same. HeyGen saves your chosen engine configuration for future generations, which helps, but your own record is what a client will ask for.

Try it

Hi, I'm Sam from the studio. Today I'll walk you through our spring range, starting with the linen jackets, and I'll show you how they wear after a week of real use.

Record this naturally as your clone sample, at least twice through. Generate the same line with two voice engines in Improve Voice, keep the closer one, then try Direct Voice set to Casual and compare it with Voice Mirror using a fresh recording.