AI voice and avatar tools price by credit, not by minute

The test

Two tools common in AI-assisted short-form production, ElevenLabs for voice cloning and text-to-speech, and HeyGen for AI avatar video, each publish a pricing page directly on their own sites. This record reads ElevenLabs' pricing page and HeyGen's pricing page as they stated their tiers on 16 September 2026.

What the evidence says

ElevenLabs lists a Free tier at '$0 per month' with '10k credits per month', a Starter tier at '$6 per month' for '30k credits', a Creator tier near $11 per month after an introductory offer with '121k credits', a Pro tier at '$99 per month' with '600k credits', and Scale and Business tiers reaching '$299' and '$990 per month' with multi-seat access. The page states text-to-speech costs '1 credit per character' and that professional voice cloning starts at the Creator tier. HeyGen's structure is similar in shape but different in unit: a Free tier at '$0/mo' capped at 'videos up to 1 min', a Creator tier at '$29/mo' with '600 credits' and videos up to 30 minutes, a Pro tier at '$49/mo' adding 4K export, and a Business tier at '$149/mo' plus a per-seat charge, extending to 60-minute videos. Both companies describe their own current retail pricing; a credit in one system is not equivalent to a credit in the other.

The sample and the variance

These are self-reported, living pricing pages, not benchmarked cost-per-output figures. ElevenLabs meters by character for text-to-speech, while HeyGen meters by a credit unit tied to 'video duration, and the complexity of the generation', so a direct per-minute or per-word cost comparison between the two tools cannot be derived from either page alone. Neither page states usage limits beyond the credit pool itself, such as maximum concurrent projects, nor do they describe enterprise-tier pricing beyond 'custom'.

What to try next

As an editorial note rather than a recommendation, a short-form producer estimating monthly cost should map expected output, script length for ElevenLabs, video minutes and complexity for HeyGen, against each platform's stated credit pool before committing to a tier, since both companies cap features like professional voice cloning or long-form export behind specific tiers rather than offering them universally. Neither company's pricing page describes consent requirements for cloning a specific person's voice or likeness in the pricing text itself, so voice cloning should not be treated as a consent-free process on the basis of these pages; that is a separate account and legal question the pricing pages do not address.

• Does the expected monthly output fit inside the credit pool of the tier being considered, given each platform's own metering unit?
• Is the feature needed, such as professional voice cloning or 4K export, actually included at the tier being priced, or gated to a higher one?
• Have either company's tier names, prices or credit allotments changed since this record was retrieved on 16 September 2026?

Both companies price access to generative audio and video tools as a metered resource rather than a flat licence, and the specific numbers here describe each company's own stated tiers as retrieved, not an independent test of output quality or cost efficiency between them.