What is a talking avatar, and how does it work?
A model maps speech to mouth and face movement and renders it onto a person. The person comes from one of three places:
- Stock avatars: licensed actors in the vendor's library. Fastest; the same faces appear in other companies' videos.
- Photo avatars: one image animated to speak. Cheapest to make, least natural over long videos.
- Custom video avatars: a few minutes of footage of a real person (you, a colleague, a hired presenter) plus their consent recording.
The voice is either the vendor's text to speech or a cloned voice; see AI voiceover for voice tools on their own. A 2-minute script is about 300 words at 150 words a minute.
Which talking avatar tools are there?
| Tool | Cheapest paid plan | Free plan | Stock avatars | Languages | Video length |
|---|---|---|---|---|---|
| HeyGen alternatives | $29/mo ($24/mo billed yearly), Creator | 3 videos a month up to 1 minute, watermark; free output is "personal, non-commercial, and internal evaluation" only | 700+ on paid plans (500+ on free) | 175+ languages and dialects | 1 min free; 30 min on Creator and Pro; 60 min on Business |
| Synthesia alternatives | $19/mo ($14/mo billed yearly), Starter | Basic: 10 minutes a month, 9 avatars, watermark | 125+ on Starter, 180+ on Pro, 240+ on Enterprise | 140+ on every plan | About 12 min a month on Starter, 60 on Pro (a per-video maximum isn't stated) |
| Colossyan alternatives | $59/mo billed yearly, Professional | Starter: 20 NEO minutes a month, videos up to 20 minutes | 300+ avatars, 700+ voices | 120+ | 20 min free; 30 min (monthly) or 50 min (yearly) on Professional |
| D-ID alternatives | $4.70/mo billed annually ($56/yr), watermarked | 14-day trial: 3 minutes, full-screen watermark | Not stated on the pricing page | Not stated on the pricing page | Not stated |
| Hedra | $20/mo, Basic (2,000 credits) | No free plan on its pricing page | Not stated | Not stated | Not stated |
Sources: HeyGen: pricing, terms, moderation policy, consent help; Synthesia: pricing, ethics; Colossyan: pricing, ethics; D-ID: pricing, Studio licence; Hedra: pricing, acceptable use. Checked November 10, 2026. Counts are the vendor's own claims. D-ID's current pricing page shows no plan amounts; its row is from our D-ID alternatives data, checked the same day.
What consent does each tool require for a custom avatar?
| Tool | Consent for a real person | Content rules |
|---|---|---|
| HeyGen | The person in a custom avatar records their own consent clip, under 30 seconds, reading HeyGen's script; their face must match the avatar footage | No avatars of real people, celebrities or public figures without explicit consent; no political opinions or campaigning; must not present output as "entirely human generated" |
| Synthesia | Consent footage "must be recorded live" with an on-screen passcode, by the same person as the avatar; photo avatars must be made by the person pictured | Stock avatars shouldn't be scripted to express personal opinions or experiences; impersonating a real person without disclosure is not supported |
| Colossyan | Can't be used to make avatars of anyone who hasn't consented; filters flag minors and public figures for human review | No avatars of politicians or public figures; "No deceptive content" |
| D-ID | You must own or have permission for anything you upload; third-party personal data needs that person's "express consent" | Lite is "non-commercial use only" under its licence; no impersonation; political use only at D-ID's discretion, with a visible watermark |
| Hedra | Content showing "individuals other than yourself" is restricted; deepfakes made "without consent or legal right" are banned | Commercial use on every paid plan; no use in elections or election campaigns |
From each vendor's terms, ethics, acceptable-use or help pages, linked above, checked November 10, 2026.
The strictest processes (HeyGen and Synthesia) make the person in the avatar record the consent themselves, so a team member can't be cloned from old webinar footage. If you're hiring a presenter for a custom avatar, put the avatar use, its duration and the right to withdraw in the contract.
What are talking avatars used for?
| Use | Why an avatar | Tools built for it |
|---|---|---|
| Training and onboarding | Update a module by editing the script, not reshooting | Colossyan, Synthesia |
| Translated versions of one video | One script, many languages with lip sync | HeyGen, Synthesia |
| Internal updates | A weekly update read by a consistent presenter | Synthesia, HeyGen |
| Help-centre and FAQ videos | Short answers to the questions support gets most | Any of the five |
| Sales and outreach | A personal-sounding intro, labelled as AI | HeyGen, D-ID |
| Social clips and ads | Presenter-led Shorts, Reels and UGC-style ads | HeyGen, Hedra (see UGC video ads) |
Where they fit less well: anything framed as a real customer's experience (see the FTC rules in UGC video ads), and software launch videos where viewers want to see the product rather than a presenter. Head-to-heads: HeyGen vs Synthesia, Synthesia vs Colossyan.
Do you have to label a talking avatar video?
- YouTube: disclose realistic content that makes a real person appear to say or do something they didn't, or generates a realistic scene that didn't occur (YouTube Help).
- TikTok ads: completely AI-generated ads need the AIGC label or your own clear disclaimer (TikTok ads policy).
- Meta ads: an AI-generated photorealistic human made with Meta's tools gets an "AI info" label next to "Sponsored" (Meta Help).
- The vendor: HeyGen's terms forbid presenting output as "entirely human generated". D-ID's licence bars removing its built-in "Synthetic Mark" without written approval.
How do you make a talking avatar video?
- Write the script for the ear, one idea per scene (150 words is about a minute). The script length calculator times it.
- Choose a stock avatar, or record footage and consent for a custom one following the vendor's steps.
- Pick the voice and language; preview the hardest line, not the first.
- Add screen recordings, slides or B-roll so the avatar isn't on screen the whole time.
- Export at the platform's size: 16:9 for YouTube, 9:16 for Reels and Shorts, and add the AI label where it applies.
Related short-form video guides
- AI voiceover: tools, prices, commercial rights and timing
- Faceless YouTube channel ideas: 32 niches with format, tools and policy risk
- Faceless Reels: formats, Instagram specs and tools
- UGC video ads: human vs AI, ad specs, FTC and platform disclosure
- Reel ideas for product launches: 25 ideas with hook, structure and length
References: Instagram Reels specs · YouTube Shorts specs · Best AI video generators · AI video models compared · Launch video script generator · Free browser video tools