To make an AI avatar video, write a short spoken script in Claude, paste it into HeyGen and choose an avatar and voice, add cutaway images from Adobe Firefly, then edit and caption the result in Descript and finish with a Canva thumbnail. Synthesia is the stronger pick for company training videos.
An AI avatar video puts a digital presenter on screen to read your script, with lip-sync and gestures. It suits training, product walkthroughs and videos you'll update often, where re-filming a person every time would be slow.
Last checked : each tool's free plan and price on the maker's own pricing page.
| What you end up with | A presenter-led video with captions, cutaway visuals and a thumbnail, ready to share. |
|---|---|
| Steps | 5 |
| Tools used | Claude, HeyGen, Adobe Firefly, Descript and Canva |
| Tools with a free plan | 4 of 5 |
| Camera or microphone needed | None |
| Coding needed | None |
One tool per step, with alternatives. Finish each step before you start the next: later steps depend on it.
Use Claude
Ask Claude for a script of about 130 words per minute in first person, written as the presenter would speak to camera. Keep sentences short: avatars sound most natural on simple phrasing.
Split the script into scenes of 2 to 4 sentences. Each scene becomes one slide or cutaway in the avatar tool.
Use HeyGen
Pick a stock avatar and voice, paste each scene into its own slide and preview before you render. Adjust pauses and the pronunciation of names and jargon.
To appear as yourself, record the short training clip the tool asks for and follow its consent steps. Never make an avatar of someone else without their written permission.
[PLACEHOLDER] Read the free plan and paid prices on heygen.com/pricing and add them as pricing.
Use Adobe Firefly
Make 1 cutaway image for each scene: a diagram, product shot or illustration of the point being made. Screen recordings work well for software walkthroughs.
Use Descript
Import the avatar video, then place cutaways over it so the presenter is on screen roughly half the time. Edit by the transcript to trim pauses and repeated lines.
Add captions and check every name and technical term in them by hand.
Use Canva
Use a frame of the presenter and 3 to 5 words that state the topic. For internal training, a clear title card works better than a clickable thumbnail.
Every price comes from the maker's own pricing page, checked on the date at the top of this page.
| Tool | Step | Free plan | Paid plans |
|---|---|---|---|
| Claude pricing (opens in a new tab) | step 1 | Free plan for chat on web, desktop and mobile, including web search and file creation. | Pro is US$17 a month billed yearly, or US$20 billed monthly. |
| HeyGen | step 2 | [PLACEHOLDER] Not checked yet | [PLACEHOLDER] Not checked yet |
| Adobe Firefly pricing (opens in a new tab) | step 3 | Free daily generations, no plan needed. | Firefly Standard is £9.98 a month including VAT, with 2,000 generative credits. |
| Descript pricing (opens in a new tab) | step 4 | Free plan for text-based editing and trying the AI tools. | Hobbyist is US$16 a month billed yearly, or US$24 billed monthly. |
| Canva pricing (opens in a new tab) | step 3, 5 | Free plan for one person. | Canva Pro is US$144 a year for one person. |
| ChatGPT pricing (opens in a new tab) | Alternative, step 1 | Free plan with unlimited chats on the standard model and limited uploads. | Go is US$8 a month; Plus is US$20 a month. |
| Synthesia | Alternative, step 2 | [PLACEHOLDER] Not checked yet | [PLACEHOLDER] Not checked yet |
| D-ID | Alternative, step 2 | [PLACEHOLDER] Not checked yet | [PLACEHOLDER] Not checked yet |
| CapCut pricing (opens in a new tab) | Alternative, step 4 | Free to use online and in the app. CapCut's website has no public pricing page, so check the Pro price in the app before you subscribe. | See the maker's pricing page |
| Adobe Express pricing (opens in a new tab) | Alternative, step 5 | Adobe Express Free costs nothing. | Premium is US$9.99 a month. |
An AI avatar is better than filming when you'll update the video often, need it in several languages or can't film to a good standard. For personal channels where viewers follow you, film yourself: an avatar can't replace your own presence.
Choose HeyGen for explainers, social videos and translating videos you already filmed. Choose Synthesia for a series of company training modules, where templates, brand settings and team seats matter more than speed. D-ID only animates a still photo, so keep it for short clips.
Whichever you pick, render one 30 second test with your own script before you commit to a subscription. Delivery quality varies more between voices than between tools.
Only make an avatar or voice of a real person with their written consent, and say clearly when a presenter is synthetic. Most avatar tools require a consent recording before they create a custom avatar. Platforms also ask you to label realistic synthetic people.
HeyGen is our pick for most people making explainers and social videos. Synthesia is better for company training and onboarding. D-ID suits quick talking-photo clips rather than full videos.
Yes. HeyGen and Synthesia both let you record a short clip to create a custom avatar of yourself, with a consent step to prove it's you. The result works best with the same framing and lighting as your recording.
Yes, when the content changes often or needs translating. You can edit the script and re-render instead of re-filming. Pair the avatar with screenshots and diagrams, because a talking head alone loses attention quickly.
Viewers accept avatars for training and how-to content, where clarity matters most. They react worse on personal or opinion channels, where they expect a real person. Label synthetic presenters clearly either way.
Not without their permission. Using a person's likeness or voice without consent can break platform rules and, in many countries, the law. Get written consent and keep a copy.
Ethics of AI from the University of Helsinki is free and covers consent, transparency and accountability, which apply directly to synthetic presenters. It needs no coding.
If you make training or explainer videos you'll update often, try this AI avatar video workflow with HeyGen; choose Synthesia when the videos are for company training, and skip avatars for personal channels.
Knowing how generative AI works makes every step above easier to get right. These are reviewed and free to start.
What large language models can actually do for you and your team, without writing a line of code. Andrew Ng teaches it in 6 hours on Coursera, certificate paid.
A free, no-code course on fairness, accountability and transparency in AI systems.
Andrew Ng's non-technical primer on what AI is good for and how to plan around it at work. It takes 7 hours on Coursera, and the certificate is paid.
Script, voice, edit and package videos faster, without handing over your style.
Script, AI voiceover, generated visuals, edit and thumbnail: a full explainer without showing your face.
The general-purpose AI you talk to. Most people should start here.
Turn a typed script into a presenter video, no camera needed.
Turn a written description into a picture.
Cut, caption and generate video with AI doing the dull parts.
Presentations, posters and graphics without a designer.
A free course on generative AI makes every prompt in this workflow better. Pick one at your level.
