HeyGen: AI Avatar Video and Video Translation – How to Use It in 2026

Section 01What HeyGen is and what it can do

HeyGen is an American service for videos with digital humans. When people talk about it, they usually mean one of two things. The first is the avatar: a person on screen looks into the camera and says any text you give it. The second is translation: you give it a finished video, and HeyGen re-voices it in another language, keeping the timbre of your voice and matching the lip movement. The viral clips where a creator suddenly speaks fluent Spanish spread across social media thanks to exactly this feature.

Over the last two years the service has grown a lot of tools. As of November 3, 2026, the pricing page lists the avatar engines Avatar III, Avatar IV, and Avatar V, HeyGen's own voice model HeyGen Voice, the video model HeyGen Video 1, and on the Pro plan also Seedance 2.5, Veo 3.1, and Nano Banana for regular video and image generation. Sounds like a zoo, but four things are enough to get started.

There are also stock presenters: 500+ ready-made avatars on the free plan and 700+ on paid plans – people of different ages and looks, in a studio, an office, or outdoors. They're handy for a first test before your own avatar is ready. For a personal brand, a stock presenter is a poor fit: your followers know your face and spot the swap right away.

HeyGen's interface is in English. Dozens of languages are available for voiceover and translation, so you can write the avatar's script in Spanish, French, or German, and the avatar will say it. Below I give the button names exactly as they appear on the site.

Section 02HeyGen for free: what you really get

HeyGen has a Free plan at $0, and it doesn't need a card. I checked the terms on the official page heygen.com/pricing on November 3, 2026:

• 3 videos a month. The counter resets once a month; unused videos don't roll over.
• Up to 1 minute each. That applies to both avatar videos and translations.
• With a watermark. The HeyGen logo sits in the frame, and only a paid plan removes it.
• One custom video avatar plus 500+ stock presenters, up to three photo avatars, and one voice clone.
• 30+ languages for voiceover and translation. Paid plans give you 175+ languages and dialects.
• Trial access to Avatar IV, Video Agent, and lip-synced translation. HeyGen's FAQ says plainly that access to these features on Free is limited and meant for trying the product before you buy.

What that means in practice. The free plan is enough to test three things: how close your twin looks to you, how your voice sounds in another language, and whether you like the result at all. For regular content it's too little: three minutes of video a month with a logo in the corner won't work for Reels.

Section 03HeyGen plans and how credits work

Since 2026, HeyGen's paid plans run on credits. Credits are a single in-app currency: you spend them on avatar videos, translation, and video and image generation. How many credits a given video takes depends on the model, the length, and the complexity of the task, and HeyGen shows the cost before you start. You can see it on the screenshot of the official page:

A few details from the FAQ on the pricing page that save money:

• Credits roll over. On monthly billing, unused credits live one more month. On yearly billing they accumulate until the renewal date.
• Credits vanish after you cancel. The subscription runs until the end of the paid period, then the account drops to Free and the remaining credits disappear. Use them up before the period ends.
• Pro comes in tiers. The Pro plan has steps: from 1,000 credits for $49 up to 100,000 credits for $4,300 a month. The features are the same; only the number of credits changes.
• Yearly billing gives you two months free. For Creator, that's $24 a month instead of $29.

Which plan to choose. If you want to make 4–8 videos a month for social media and translate a video now and then, start with Creator. You need Pro in two cases: you translate long webinars and lessons and want to proofread the translated script before voicing, or 4K matters to you. Business is for online schools where several people work on the videos.

Section 04How to sign up for HeyGen

Signing up takes about a minute, and you don't need a card for the free plan. Here's the order that works.

Section 05Your first video with a stock avatar: step by step

Start with a stock presenter. That way you'll learn the editor in ten minutes and see how HeyGen reads text, and only then spend time on your own twin.

Section 06Digital twin: an avatar made from your video

The digital twin is the main reason people who sell their knowledge come to HeyGen. You record yourself on video once, and then for months you publish videos from text. 90% of the twin's quality depends on the source recording, so I'll go through it in detail.

How to record the source video

• 2–5 minutes in one take. No cuts, edits, or transitions. Say anything: talk about your work or read any text. The model needs to see how your lips move on different sounds.
• Eyes on the lens. Look straight into the camera lens. If you look at the phone screen, your twin will look slightly past the viewer for the whole video.
• Light in front. The window or lamp is in front of you, with no light behind you. Shadows on half your face get carried into every future video.
• A quiet room and a lav mic. HeyGen takes your voice print from the same recording. The echo of an empty room and the hum of the fridge become part of your digital voice too.
• Pauses with your mouth closed. Between phrases, go quiet for a second and close your mouth. Then your twin can stay silent naturally instead of freezing with its mouth open.
• Calm hands. Keep your hands below your chest, and don't touch your face or hair. Gestures near the face cause artifacts: fingers blur and pass through the chin.
• Clothes and background without fine stripes or checks, and no green if you plan to replace the background. Wear what you want to appear in for every video over the next few months.

How to create your twin in HeyGen

Open Avatars, click Create avatar, and choose the video option. Upload the recording. Next, HeyGen asks for a consent video: you record yourself again and read a short text aloud saying you allow the avatar to be created. That's how the service guards against people who want to copy someone else's face. You can't make a twin of a well-known creator from their social videos, and that's the right call.

Processing takes anywhere from a few minutes to a couple of hours. Once the twin is ready, it shows up in your avatar list, and from there it works just like the stock presenter from the previous section: pick yourself, paste the script, click Generate. The free plan gives you one custom video avatar; on Creator you build an avatar with your own voice and look and export videos in 1080p with no watermark.

Section 07A talking video from one photo

A photo avatar is the fastest way to get a video with your face. You upload one portrait, add a script or audio, and HeyGen makes a video where the person in the photo talks, blinks, turns their head, and gestures. The Avatar IV engine handles this, and you can try it even on the free plan.

How to do it:

1. Open Avatars and choose to create a photo avatar.
2. Upload a photo: face straight on or at a three-quarter angle, eyes open, mouth closed, no sunglasses, no hands near the face. The larger the face and the less compressed the photo, the fewer artifacts.
3. Paste the script and choose a voice. If you have a clone of your own voice, pick it. You can also upload ready audio, and the lips will follow the recording.
4. For Avatar IV you can set gestures and mood with a short description, for example calm, friendly, small hand gestures. The free plan gives you 3 of these gesture controls a month, Creator gives you 20.
5. Click Generate and check how the teeth, ears, and hair behave. Those are the places where a photo avatar gives itself away most often.

Where a photo avatar shines: short videos under a minute, answers to questions in Stories, live-stream announcements, a greeting in a chatbot. Where it loses to a twin: long lessons. By minute five, viewers notice the head cycling through the same few movements. For a course, record a video twin. If you just want an old photo to move without speech, see how to animate a photo with AI.

Section 08Translating a video into another language in your voice

Translation is HeyGen's second star feature and, in my view, the best value. You don't reshoot anything. You take a webinar, lesson, or Reel that already worked in English and get it in Spanish, German, Portuguese, or another of 175+ languages. The voice keeps your timbre, and the lips move to the new words.

Three tips that make the translation sound alive

• Remove music under the voice. HeyGen separates the voice from the background, but loud music still creeps into the translation as crackle. If you have a version without music, translate that, and add the music back later in your editor.
• Speak shorter in the original. Spanish and German usually run longer than English. To fit the timing, HeyGen speeds up the speech, and on screen you start rattling. If you plan to translate, write the script in short phrases with pauses.
• Set up a glossary. Paid plans include a brand glossary: you add product names, people's names, and terms that must not be translated. For example, the name of your course.

Section 09Let Claude write the avatar's script

The weakest spot in avatar videos is the text. Paste a paragraph from a post into HeyGen, and the avatar will read it evenly and dully, like a train station announcer. A script for voiceover is written differently: short phrases, one thought per phrase, pauses, conversational words. This is an easy job to hand to Claude. If you haven't worked with it yet, start with Claude from scratch; you can sign up at claude.ai.

A prompt you can copy and adapt:

What worked: the question hook in the first line, short phrases the avatar says without stumbling, and concrete actions instead of general advice. What I'd fix before loading it into HeyGen: "Plus, you're tired" is better split into two phrases with a pause, or the avatar swallows the transition. You don't need to paste the "/" marks into HeyGen; put a pause in the editor in their place. And time it yourself with a stopwatch: 140–150 words in English is roughly a minute of calm speech.

Two more prompts that come in handy:

• Prep for translation: "Here's the transcript of my video in English. Rewrite it 15% shorter, keep the meaning and my words, and split long sentences into short ones. The video will be translated into Spanish, and the speech must not speed up."
• A series of videos: "Here are 10 common questions from my clients. Turn each one into a 40-second avatar script: a hook, the answer in three steps, and an invitation to send me a direct message. One topic per video."

More ready prompts for video are collected in AI video prompts, and ideas for a whole month of videos are in AI for Reels: 30 scripts in one evening.

Section 10Avatar video ideas for different niches

An avatar shines wherever you need many similar videos with the same person in the frame. Here's where it saves the most time:

Where you don't need an avatar: personal stories, behind the scenes, emotions, and arguments in the comments. There viewers want to see the real you, mistakes and laughter included. A good split: the avatar for useful and teaching videos, live video for everything personal. You can combine the avatar with live footage, graphics, and captions in any video editor. For other ways to make video from a text prompt, see AI video from text.

Section 11If something goes wrong

Section 12HeyGen, ElevenLabs, and AI video tools: what's for what

HeyGen often gets compared with services that solve neighboring tasks. In practice they complement each other well.

A working combo looks like this: Claude writes the script, HeyGen makes the video with your twin, Kling or Hailuo provide 3–5 second B-roll shots, and you do the final cut with captions in your editor. One video like that comes together in 20–30 minutes, and you don't have to set up lights or sit in front of a camera.

Section 13The whole path in one minute