HeyGen Voice Turbo Is Free in the API Until Nov 30
This page contains one affiliate link, to ElevenLabs, clearly marked where it appears — HeyGen has no affiliate relationship with us and this free offer is theirs, not ours. We have not run production traffic through HeyGen Voice Turbo yet — this is researched analysis of HeyGen's announcement and the Artificial Analysis leaderboard, not hands-on testing. Where a number is HeyGen's own marketing claim, we say so.
The top-ranked text-to-speech model on the Artificial Analysis leaderboard is free through the API right now. No credit card, $0 per million characters, and the only catch is a deadline: the free window closes November 30, which means eight days are left as we write this.
HeyGen Voice is the in-house voice model behind every HeyGen avatar video. Turbo is its fast mode: same voices, shorter generation step, audio back sooner. Faster usually means worse, but in the blind listening tests behind the Artificial Analysis text-to-speech leaderboard, Turbo ranks first and the full HeyGen Voice model ranks second. Both sit ahead of ElevenLabs' Eleven v4 Turbo and Eleven v4.
The leaderboard setup matters here. Every model reads the same eight cloned voices, four US and four UK, and listeners pick winners blind, two clips at a time. So the ranking is about the model, not about one provider having a nicer demo voice. That's a fairer fight than most vendor benchmarks. HeyGen also claims its users preferred Turbo over ElevenLabs 64% of the time in blind side-by-side listening — that's HeyGen's number, from HeyGen's users, so weigh it accordingly. The stronger signal is the leaderboard rank, not the 64%.
🔍 How we put this together
We used HeyGen's announcement "HeyGen Voice Turbo Is Free for Developers" (published October 21, 2026) and the Artificial Analysis text-to-speech leaderboard it cites. Pricing, the free-tier limits and the November 30 end date are all from HeyGen's own post. No hands-on testing was performed for this article.How to use the free voice API
Create an API key on platform.heygen.com, then call the text-to-speech endpoint with the free model name. You get back a link to a WAV file:
curl -X POST "https://api.heygen.com/v3/models/audio/tts" \
-H "X-Api-Key: $HEYGEN_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "heygen-voice-1-turbo-free",
"voice_id": "83548de556df48ba8c09c42a57c51d85",
"text": "Hello from HeyGen Voice Turbo."
}'
That's the whole integration. If you're already calling a text-to-speech API somewhere, this is an afternoon's experiment, not a migration. There's also a playground on platform.heygen.com if you want to hear the public voices before writing any code.
What "free" actually means
Free tiers always have fine print, and this one's is refreshingly readable:
- Best effort only. No uptime or latency guarantee. When free capacity fills up, requests return a 429 and you retry later. Fine for prototypes and evaluation. Not fine for a live product your users depend on.
- Public voices and instant clones only. You can use HeyGen's public voices and clone an instant voice from a single recording, and creating instant voices is free during this preview too.
- Professional voice clones are not included. Those need the paid full model,
heygen-voice-1. - It ends November 30, 2026. HeyGen says it will give advance notice before anything changes, but plan around the date, not the promise.
What it costs after November 30
Switching to production is a model-name change — nothing else in the request moves:
| Model | Price | Notes |
|---|---|---|
| Turbo Free | $0 until Nov 30 | Best effort, public + instant voices |
| Turbo | $4 per 1M characters | Fast, priority |
| HeyGen Voice | $15 per 1M characters | Standard, includes professional clones |
For scale: a million characters is roughly 18 hours of finished narration. Four dollars for that, at the quality the leaderboard suggests, is aggressive pricing by any measure in this market.
Where this pressures ElevenLabs, and where it doesn't
The pressure is real on two fronts. First, the leaderboard: ElevenLabs has owned the "best-sounding API voice" conversation for a long time, and it is currently ranked below both HeyGen models in the most controlled public comparison we have. Second, price: a flat $4 per million characters for the fast tier undercuts the credit-plan math most developers do with ElevenLabs.
Where ElevenLabs and Murf still make sense today:
- Production guarantees. The free Turbo tier is explicitly best effort. ElevenLabs sells a service with actual plans, limits you can plan around, and a status page that means something. If voice is in your product's critical path this week, that difference is the whole ballgame.
- Professional voice cloning. A trained, high-fidelity clone of a specific voice is a paid HeyGen feature, and it's the thing ElevenLabs built its reputation on. If your brand voice is a professional clone, nothing about this announcement changes your stack today.
- Ecosystem depth. Conversational agents, dubbing, sound effects, a mature dashboard and years of integrations. HeyGen's voice API is new; ElevenLabs' platform is not. Murf remains a solid pick for studio-style narration workflows for the same reason: maturity.
If you do need that production-grade route today, ElevenLabs is still the safer default in our book. Full disclosure: this is our affiliate link to ElevenLabs — it costs you nothing extra, and it's the only affiliate link in this piece.
What we'd do with the eight days
Run your real scripts through Turbo, not demo sentences. Compare the same text on your current stack, blind if you can manage it, and write down where each one wins. Clone an instant voice while that's free too, because that's the feature most likely to change how you think about the tool. And put November 30 on the calendar: if Turbo earns a place in your stack, the switch to the $4 tier should be deliberate, not a surprise invoice.
Sources: HeyGen, "HeyGen Voice Turbo Is Free for Developers," published October 21, 2026 (https://www.heygen.com/blog/heygen-voice-turbo); Artificial Analysis text-to-speech leaderboard (controlled voice), as cited in the announcement. The 64% preference figure is HeyGen's own claim from its own users. No hands-on testing was performed for this article. Affiliate disclosure: this article contains one affiliate link, to ElevenLabs, marked where it appears. HeyGen has no affiliate relationship with this publication.