How to clone your voice free with HeyGen Voice

What changed on October 9

On October 9, 2026 HeyGen announced HeyGen Voice, an in-house voice model built, in the company's words, to keep your tone, pacing and expression. HeyGen's stock and designed voices run through third-party speech providers; this is the first model HeyGen built itself. HeyGen says the new model debuted at number one on Artificial Analysis' voice leaderboard.

• HeyGen Voice is free inside the HeyGen platform and API.
• HeyGen Professional Voice Clone is a $99 per month add-on. It is trained, with your explicit consent, on 30 minutes to three hours of your speech, for a closer match to the real you.
• HeyGen requires the voice owner's explicit consent before a voice is cloned.

The announcement also quoted a HeyGen survey of more than 1,000 small business owners: 71.6% had recorded a business video and never posted it.

What a voice clone does

You upload one recording of yourself. HeyGen builds a voice from it. From then on you type a script and the clone reads it, in your voice, in any of the supported languages. You never have to record that script yourself.

• Instant clone: one recording, ready within seconds, free. HeyGen's developer docs say only the first three minutes of the recording are used, and an instant voice cannot be retrained (to change it, make a new one).
• Professional clone: the $99 per month add-on, trained on a longer set of recordings for a closer match. HeyGen's announcement says 30 minutes to three hours of speech; the developer docs accept from 20 minutes across up to ten recordings.
• Languages: cloning and speech support 39 base languages, including English, Spanish, French, German, Japanese and Chinese. A regional tag such as en-US is accepted but does not change the accent.

Check you have what you need

• A HeyGen account. The Free plan lists 1 voice clone and 3 videos a month (up to one minute each). Paid plans start at $29 a month for the Creator plan with expanded usage.
• A quiet room and a decent mic. HeyGen's own guide says to use an external mic, a Bluetooth mic or a recent phone, and to avoid the laptop's built-in mic.
• One to three minutes of you talking naturally. HeyGen's tool page says a 30-second recording is enough for an instant clone; its developer docs suggest one to three minutes of clear speech and use only the first three minutes.
• Permission. Clone your own voice, or a voice whose owner has agreed in writing. HeyGen requires explicit consent.

Clone your voice in HeyGen

1. Open HeyGen and open a project in AI Studio (or create one).
2. In the script panel, click the current voice to open the voice menu, then select Add New Voice.
3. Choose Create New Voice. (The other option is to connect a third-party voice provider.)
4. Upload your recording. HeyGen generates the voice clone.
5. Listen. If it is not right, click Enhance Voice: adjust the accent or voice engine, or describe what sounds off, and Voice Doctor generates options you can preview.
6. Save the version you like as a new voice, or replace the one you have.

You can also manage voices from the Avatars page. If you would rather not record at all, Voice Design builds a voice from a text description instead.

Record a sample worth cloning

The clone copies whatever you give it, so the sample matters more than the settings.

• Record in a quiet room, phone or external mic close to your mouth, no music or fan in the background.
• Speak the way you talk to a customer, not the way you read. Vary your emotion a little; HeyGen's guide asks for natural speech with some emotional variation.
• Keep it to one to three minutes. Anything past three minutes is ignored on an instant clone.
• Use MP3 or WAV. OGG works only from a public URL.
• Read a real script, for example your usual phone greeting followed by a short explanation of a service.

Make it read your script

Once the voice is active, write or paste your script in the script panel and pick your clone as the voice. For an avatar video, generate as usual. For audio only, HeyGen's AI Voice Generator lets you render the voice-over and download it as an MP3 without rendering a full avatar video: enter the script, choose the voice, adjust pace and emphasis, then generate and download.

A script that works for a 30-second business update:

Hi, I'm Dana from Harbor & Pine.
Quick update this week: we now offer same-day quotes.
Send us a photo of the job, and you'll have a price by 5 pm.
Book at harborandpine.com or just reply to this message.

Short sentences read better than long ones. Spell out numbers you want said a particular way ("five p m", not "5pm"), and put a full stop where you want a pause.

Instant or professional?

Start with the free instant clone. Move to the professional clone only if you will use the voice every week and the instant one does not sound like you to the people who know you.

Limits to know

• The Free plan lists one voice clone. A failed clone stays in your list until you delete it.
• An instant clone cannot be retrained. Record a better sample and make a new voice.
• HeyGen's developer docs cap one speech request at 5,000 characters, so split a long script into parts.
• HeyGen says it requires the voice owner's explicit consent. Only clone yourself, or someone who has agreed.
• Tell your audience when a video uses a clone if they would reasonably assume you recorded it live. It costs nothing and keeps trust.

Questions people also ask

Sources

• HeyGen launches HeyGen Voice, debuting at #1 on Artificial Analysis' leaderboard (press release, October 9, 2026)
• HeyGen Academy: How to create a new voice
• HeyGen developer docs: HeyGen Voice instant clone
• HeyGen developer docs: HeyGen Voice speech
• HeyGen developer docs: Models
• HeyGen pricing
• HeyGen: AI voice cloning tool page
• HeyGen blog: Generate audio only in HeyGen