⚖️ Last reviewed May 2026 · 15 providers · 50+ voices
Which text-to-speech voices can you actually use commercially? We parsed the real terms of service for 15 major providers. Here's what each one actually allows for YouTube, podcasts, audiobooks, SaaS, IVR, and resale.
What each column actually means
Every column in the matrix below answers a specific yes-or-no question. Here's what each one means in plain terms.
Commercial usemeans using the audio in any context where money changes hands, directly or indirectly. YouTube monetization, selling an audiobook, using TTS in a paid app, running ads on a podcast — all commercial. If there's revenue anywhere in the chain, it counts.
YouTube monetization means running ads on videos that include the generated audio. This is usually the same as commercial use generally, but some providers have addressed YouTube specifically in their docs.
Resell in products means selling something that contains the audio. An audiobook. A course. A game. An app. This is distinct from selling access to the TTS service itself, which no provider allows without a special reseller agreement.
Voice cloningmeans creating a synthetic replica of a specific person's voice. Where it's marked Conditional, it means the feature exists but requires explicit consent from the voice subject. Where it's No, either the feature doesn't exist or it's outright prohibited.
Free tier commercial answers whether commercial rights carry over to the unpaid tier. For cloud APIs, usually yes. For SaaS tools, usually no.
Attribution required means the ToS requires you to credit the provider in your content or product.
15 providers, 6 rights dimensions each
| Provider | Commercial use | YouTube monetized | Resell in product | Voice cloning | Free tier comml. | Attribution req. | Plan |
|---|---|---|---|---|---|---|---|
| FreeTTS.org | Yes | Yes | Yes | Yes | Yes | No | Free & Pro |
| Google Cloud TTS | Yes | Yes | Yes | Conditional | Yes | No | All plans |
| Amazon Polly | Yes | Yes | Yes | Conditional | Yes | No | All plans |
| Microsoft Azure TTS | Yes | Yes | Yes | Conditional | Yes | No | All plans |
| OpenAI TTS | Yes | Yes | Yes | No | Yes | No | All plans |
| ElevenLabs | Paid only | Paid only | Paid only | Conditional | No | No | Creator+ only |
| PlayHT | Paid only | Paid only | Paid only | Conditional | No | No | Paid plans |
| Murf.ai | Paid only | Paid only | Paid only | Conditional | No | No | Pro+ |
| WellSaid Labs | Paid only | Paid only | Paid only | Conditional | No | No | Paid plans |
| Speechify API | Paid only | Paid only | Paid only | Conditional | No | No | Business+ |
| TTSMaker | Yes | Yes | Yes | No | Yes | Yes | Free (with attribution) |
| Replica Studios | Paid only | Paid only | Paid only | Conditional | No | No | Paid plans |
| LOVO.ai | Paid only | Paid only | Paid only | Conditional | No | No | Pro+ |
| Coqui TTS | Conditional | Conditional | Conditional | Conditional | Conditional | Conditional | Model-dependent |
| Bark (Suno AI) | Yes | Yes | Yes | No | Yes | No | MIT License |
Last verified May 2026. ToS links for each provider are in the deep-dive sections below.
What the actual ToS says, in plain language
FreeTTS allows commercial use on all tiers. Generate audio on the free plan, put it in your YouTube video, run ads. Generate on a Pro plan, sell the audiobook. No attribution required in either case.
Voice cloning is available on Pro and covers your own voice. You generate it, you own it, and there are no hidden clauses restricting what you do with it commercially. The usual content restrictions apply: nothing illegal, nothing designed to deceive.
Google treats your TTS output as your customer data under the Cloud Customer Agreement. That means commercial use, YouTube monetization, and resale in your products are all standard. The free tier is a billing promotion, not a different license — same Cloud ToS, same commercial rights.
The Voice AI API (Google's custom voice feature) requires a separate agreement and proof that you have rights in any training data used. Standard pre-built voices? No extra paperwork. No attribution required in your content. The Cloud AUP still applies: no illegal uses, no deceptive content in regulated contexts.
Polly operates under the AWS Customer Agreement. Generated audio is your customer content. Podcasts, audiobooks, YouTube videos with ads, apps, IVR systems — all standard commercial use and all allowed. The AWS 12-month free tier carries the same agreement, so you get commercial rights from day one.
The one thing you can't do is build a TTS platform that exposes Polly as your own voice engine and sell access to it. That's API resale, and it requires an AWS partner agreement. Selling your own products that contain Polly-generated audio is fine. Brand Voice (custom voice training) needs a separate contract and documented consent from the voice talent.
Azure Neural TTS is commercial by default under Microsoft's Online Services Terms. Standard plan covers commercial use, YouTube, and product resale without attribution. The free tier operates under the same OST.
Custom Neural Voice (CNV) has a strict separate policy. Microsoft requires explicit written consentfrom the voice talent, and their Responsible AI standard bars deepfakes and political impersonation. Some voices are "Limited Access" — you have to apply for them and Microsoft approves use cases individually. For standard pre-built voices, none of that applies. No attribution required.
OpenAI is one of the clearest on this. Their Terms of Use directly address commercial output: you can use content generated by OpenAI — including audio from the TTS API — for commercial purposes including sale or publication. No ambiguity, no plan restriction.
The main restriction is impersonation. OpenAI's Usage Policies prohibit using TTS to impersonate real people without their consent. There's no general voice cloning feature in standard TTS — you get six preset voices. What you can't do is use those voices in a way that makes people think a real person said something they didn't. No attribution required.
ElevenLabs draws a hard line between personal and commercial use. The free tier is explicitly personal use only. If you've been generating audio on the free plan and putting it in monetized content, you're outside their ToS. The Creator plan is where commercial rights begin.
On paid plans, YouTube monetization and audiobook resale are allowed. Voice cloning is a core ElevenLabs feature, but the rules are strict: you need explicit consent from the person you're cloning, and you absolutely cannot clone a third-party celebrity or public figure's voice without their permission. ElevenLabs has been aggressive about enforcing this, including account terminations for violations.
Their plan structure has changed multiple times. Always verify the current pricing page for what commercial rights are included at your tier.
PlayHT markets heavily to creators and content producers, so commercial use is the whole point of their paid plans. On any paid tier, you can use generated audio in monetized YouTube videos, podcasts, sold audiobooks, and paid apps. The free plan is for testing the product, not for production content.
PlayHT 2.0 includes voice cloning. Same consent rules apply: your own voice, fine. A third party's voice, you need their documented consent. No selling raw voice clones without an OEM agreement. And you can't resell API access as your own competing TTS service.
Murf uses "royalty-free license" language for generated content on paid plans. That covers YouTube monetization, selling e-learning courses, explainer videos, ads, and similar productions. The free plan is evaluation only — no commercial use.
Some voice actors in the Murf library have additional restrictions, particularly around political advertising. If you're creating political content at any level, check the individual voice notes in the Murf interface before using that voice in production. Custom voice features require consent and a separate agreement.
WellSaid is built for enterprise and professional studio use. Commercial rights on paid plans cover corporate training videos, YouTube, ads, and product narration. They don't have a real free tier — trial access exists for evaluation but carries no commercial rights.
Custom Voices come with a dedicated contract and require strict written consent from the voice talent. Some WellSaid voices have restrictions around political content. No attribution required for standard commercial use.
Speechify's main ToS states directly that their Business, Enterprise, Studio, Voice-Over, and API offerings are intended for commercial use. The consumer reading app has no commercial API rights. If you're building on Speechify's voice generation for production use, you need one of those commercial product lines.
The AI Voice API has a supplemental terms document worth reading separately from their main ToS. It covers data handling, output rights, and voice cloning consent requirements in more detail than the main agreement.
TTSMaker stands out by offering commercial use on the free tier, which is genuinely unusual in this space. The catch is attribution: you must credit TTSMaker in any commercial content using their audio. If your product or brand doesn't want a visible attribution to a third-party TTS tool, this won't work for you.
No voice cloning on the free tier. Paid plans may relax attribution requirements — check the current pricing page for details. For anything where clean, unbranded audio is important (branded IVR, professional narration), you'll want a cloud API or a paid SaaS plan instead.
Replica Studios targets game developers and entertainment productions. Commercial rights on paid plans cover games, films, YouTube, and similar media. They have explicit licensing language around game development use cases, which is more detailed than most competitors.
Voice cloning is allowed with consent, but they're particularly strict about character likeness: if you're trying to replicate a well-known game character or actor's voice without IP rights, that's a hard no. Their consent and likeness documentation is worth reading if you're doing entertainment work.
LOVO markets to YouTube creators, e-learning producers, and ad agencies. Commercial rights on paid plans cover those use cases. The free plan is evaluation only. Standard SaaS consent rules apply for voice cloning. No attribution required on paid plans.
Coqui TTS is a family of open-source models, not a single product. Each model checkpoint has its own license, and they vary significantly. This is the most important thing to understand before building anything on Coqui.
XTTS v2, which is the flagship voice cloning model, uses the Coqui Public Model License (CPML). That license restricts commercial use for companies with more than $1 million in annual revenue or more than 10,000 end users. Below those thresholds, commercial use is generally allowed. Above them, you need a separate commercial license from Coqui.
Other Coqui models use Apache 2.0 or MIT licenses, which are fully permissive for commercial use. Read the model card — not the repository README — for the specific checkpoint you're using. Don't assume the repo-level license applies to every model in it.
Bark's code is MIT licensed. MIT is one of the most permissive software licenses: you can use, modify, distribute, and sell products built on it commercially without paying anything or requiring attribution. That covers YouTube monetization, audiobooks, apps, and SaaS products.
The model weights should be checked separately from the code license — these sometimes differ. Verify the current model weights license on the Suno Bark GitHub repository before large-scale commercial deployment. Bark doesn't offer traditional voice cloning in the way ElevenLabs or XTTS do — voice consistency across generations is more constrained.
Which tools work for your specific situation
Running ads on YouTube counts as commercial use. All cloud APIs (Google, Polly, Azure, OpenAI) cover this on all plans. ElevenLabs, Murf, and PlayHT cover it on paid plans. TTSMaker covers it free with attribution. FreeTTS.org covers it on all tiers.
Monetized podcasting is commercial use. Any tool that allows YouTube monetization covers podcasting too. Patreon revenue from a podcast containing AI audio is the same as any other commercial content.
Selling an AI-narrated audiobook requires a TTS license that covers resale in products. Cloud APIs cover this on all plans. ACX (Audible) requires you to disclose AI narration at submission. KDP and Draft2Digital have similar disclosure requirements.
Embedding TTS in a sold or subscription app is commercial resale of a product that contains TTS audio. Cloud APIs allow this by default. SaaS tools usually allow it on paid plans. The thing you can't do is resell API access as a competing TTS service without a reseller agreement.
Broadcast commercial use is generally covered under the same commercial rights as other content, but some voice actors in SaaS libraries have specific restrictions on political advertising or broadcast at scale. Check voice-level terms for anything going on broadcast media.
IVR (Interactive Voice Response) systems are a primary commercial use case for cloud TTS APIs. Google, Polly, and Azure all have dedicated IVR features and pricing. Dynamic prompt generation for high-volume call centers is where API-based providers have a clear advantage over SaaS tools.
What consent means and why every provider requires it
Voice cloning is the most legally sensitive area of TTS right now. Every major provider that offers it requires explicit consent from the voice subject. That's not a coincidence — it's because the legal exposure from unauthorized cloning is substantial.
Most U.S. states have right-of-publicity laws that give individuals the right to control commercial use of their likeness, including their voice. California's law is among the most protective, covering digital replicas. Illinois has BIPA (Biometric Information Privacy Act) with teeth that allow private lawsuits. The EU has GDPR protections around biometric data. In 2024 and 2025, multiple states passed or proposed laws specifically targeting AI voice cloning.
For practical purposes, here's what consent means across the major providers:
Microsoft Azurerequires written consent — actual documentation. "Explicit written consent" is the standard phrase. That means a signed agreement, not just a verbal OK.
ElevenLabs requires you to certify that you have permission when you upload a voice sample. Their platform has voice similarity detection and will flag voices that appear to be public figures. Accounts violating this have been terminated.
PlayHThas similar consent certification in their voice upload flow. They also prohibit using voice clones to impersonate people in a way that could deceive others about who's speaking.
OpenAI TTSdoesn't offer arbitrary voice cloning — you use their preset voices. But their usage policy explicitly prohibits using those voices to impersonate real individuals without consent.
You own your voice. Cloning and commercially using your own voice is legal and covered by every provider that offers cloning. There's no special agreement needed beyond your account ToS. Make sure you're the only one with access to the clone, and that your commercial use case falls within the provider's allowed uses.
Clone a recognizable celebrity or public figure without their documented, contractual permission. Use a voice clone to put words in someone's mouth in a way that could deceive viewers. Use a cloned voice in political advertising without specific compliance review. Create synthetic audio designed to defraud, harass, or defame.
And one practical note: even if a platform allows it technically and you have consent, the major distribution platforms (YouTube, podcast hosts, Audible, app stores) have their own policies on synthetic media that you need to comply with separately.
Notable TTS policy changes in 2025 and 2026
Several U.S. states passed legislation specifically targeting AI voice replication, requiring explicit consent for voice cloning and establishing civil liability for unauthorized use. If you operate commercially in the U.S., check the current status of voice cloning laws in your state and your users' states.
ElevenLabs updated their plan structure and commercial rights language multiple times through 2025. The naming of plans that include commercial rights changed. Always verify on their current pricing page which plan you need for commercial use rather than relying on third-party summaries.
OpenAI updated their usage policies with more specific language around synthetic audio, impersonation, and voice use in political contexts. The commercial-use rights for output remained intact, but the prohibited uses section expanded.
Microsoft expanded the set of voices requiring individual application and approval under their Limited Access program. Standard pre-built voices were unaffected for commercial use, but custom voice deployments now have stricter review requirements in certain verticals.
Coqui released their XTTS v2 voice cloning model under the Coqui Public Model License rather than a fully open license, creating commercial-use thresholds for the first time in their model lineup. This caught some commercial users off guard who assumed all Coqui models used permissive open-source licenses.
Before you ship any commercial TTS content
17 questions about TTS commercial use, answered plainly
No plan upgrade required. Generate audio, use it commercially, keep it. No attribution. No logs.
Start generating free →