Text-to-speech has quietly become one of the most competitive corners of the AI API market. Voice quality has improved, response times have dropped sharply for models built for real-time use, and pricing covers a wide range depending on what you're optimizing for. This roundup compares the major text-to-speech APIs on the things that actually affect a build decision: price, voice cloning support, speed, and where each one fits best.
If you're also working with AI-generated video, music, or images, it helps to treat text-to-speech as part of a larger media pipeline rather than a standalone feature. Apiframe's AI Video Generation API guide explains how developers build similar media workflows through a single API.
What to Look for in a Text-to-Speech API
Voice quality and naturalness. This one is subjective, so test it with your own scripts instead of trusting a provider's demo reel, since demos are chosen to flatter. Pay attention to how a voice handles pauses, emphasis, and uncommon words, not just short, clean sentences.
Speed: real-time vs batch. If you're building a voice agent or anything conversational, the delay before the first audio reaches the listener matters a lot, and the gap between providers is wide. If you're generating narration for a video or podcast in advance, that delay barely matters.
Voice cloning support. Some providers can clone a voice from a short sample in minutes, others require more setup or a verification step, and some don't offer it at all. This is also where consent and licensing questions become real, which we cover below.
Language and accent coverage. This varies a lot by provider. If you need strong non-English support, check it specifically rather than assuming a provider that sounds great in English will be just as good elsewhere.
Licensing for commercial use. Read the terms before you build. Some providers limit how cloned voices can be used commercially, and that limit usually sits on a licensing page most people never open until it causes a problem.
Pricing model. Per-character, per-minute, and monthly subscription plans all show up. They aren't directly comparable until you convert them to the same unit, which the table below does for you.
Quick Comparison Table
| Provider | Starting price | Voice cloning | Latency | Best for |
|---|---|---|---|---|
| ElevenLabs | ~$50/1M chars (Flash), ~$100/1M chars (Multilingual v2) | Yes, instant | Low (Flash tier) | Voice cloning quality, multilingual |
| OpenAI TTS | ~$15/1M chars (standard), ~$30/1M chars (HD) | No | Moderate | Simplicity, OpenAI ecosystem fit |
| PlayHT | ~$30-40/1M chars | Yes | Moderate | High-volume narration, audiobooks |
| Cartesia | ~$5-37/1M chars depending on plan | Yes | Very low (~40ms) | Real-time conversational voice |
| Google Cloud TTS | Varies by model tier | Limited | Low | Language and accent breadth |
| Amazon Polly | ~$4/1M chars (standard voices) | No self-serve | Low | Enterprise, AWS-native billing |
| Deepgram Aura-2 | $30/1M chars (pay-as-you-go) | No | Low | Developer-first, API-only positioning |
Prices change often, so check each provider's current pricing page before you commit. The relative ordering here is a reasonable starting point.
The Best Text-to-Speech APIs
ElevenLabs
ElevenLabs is widely seen as the leader in voice cloning, with instant cloning from a short sample and consistently high naturalness scores in independent comparisons. It sits at the higher end of the market: Multilingual v2 runs about $0.10 per 1,000 characters, more than three times Deepgram's rate. The Flash v2.5 model brings that down to roughly $0.05 per 1,000 characters if you don't need the top-tier model for every request. Apiframe's ElevenLabs API guide covers pricing and setup in more depth, and Apiframe also offers ElevenLabs Music through the same endpoint used for its other audio and video models.
OpenAI TTS
If you're already building on OpenAI, its text-to-speech models are the easiest option to add: same API key, same SDKs, and simple pricing at roughly $15 per million characters for the standard tier and $30 per million for the higher-quality tier. It doesn't lead on any single measure, and it doesn't support voice cloning, but that simplicity is the real selling point.
PlayHT
PlayHT focuses on real-time streaming and a large voice library of more than 900 voices, with pricing in the $30 to $40 per million character range on its higher-volume plans. It's a good fit for high-volume narration work such as audiobooks and long-form podcasts, where having plenty of voices to choose from matters more than shaving milliseconds off response time.
Cartesia
Cartesia's Sonic model is built for real-time use, with a quoted delay of about 40 milliseconds before audio starts playing, among the fastest in the category. Pricing is credit-based and works out to roughly $5 to $37 per million characters depending on the plan. If you're building a voice agent where someone is waiting for a reply, that speed is the main reason to choose Cartesia over a slower option at a similar price.
Google Cloud Text-to-Speech
Google's strength is breadth: more languages and regional accents than most competitors, backed by its translation and language infrastructure. Pricing varies by voice tier, and Google offers several, so check the current rate for the specific tier you need rather than assuming one number applies across the board.
Amazon Polly
Polly is the obvious choice if you're already running on AWS, both for the built-in billing and for enterprise purchasing reasons. Pricing is around $4 per million characters for standard voices, making it one of the cheaper options here, though it trails newer entrants on naturalness and has no self-serve voice cloning. AWS does offer custom Brand Voice work, but that runs as a separate engagement rather than something you can set up from the API.
Deepgram Aura-2
Deepgram built Aura-2 as a developer-first API rather than a consumer app with an API added on top. Pricing sits at $30 per million characters on pay-as-you-go and drops slightly on higher-volume plans. It's a reasonable middle ground on price and speed if you don't need voice cloning and want a straightforward, well-documented API.
Text-to-Speech API Pricing Models Explained
Per-character pricing, usually quoted per 1,000 or per million characters, is the most common model and the easiest to budget for, since you know how long your text is before you send the request. Per-minute pricing is less common for text-to-speech and only becomes comparable once you know how many characters you produce per minute of speech (roughly 800 to 1,000, though this varies by language and speaking rate). Subscription plans bundle a fixed character allowance into a monthly fee, which usually costs less per character than pay-as-you-go if your volume is steady, but wastes money if it isn't.
To estimate your monthly cost, take your expected audio output in minutes, multiply by about 900 characters per minute, then multiply by the provider's per-character rate. A podcast producing 10 hours of narration a month works out to roughly 540,000 characters, which sits well within most providers' starter or pay-as-you-go tiers.
If you're already working with other AI media APIs, Apiframe's AI Music API guide takes a similar approach to comparing pricing, licensing, and integration requirements for AI-generated audio.
Voice Cloning: What to Check Before You Build
Voice cloning raises consent and licensing questions that are worth settling before you ship a feature, not after. Reputable providers ask for proof that the person being cloned has agreed to it, and several now attach origin data to generated audio so it can be traced back to its source. In the US, the FCC ruled in February 2024 that calls using AI-generated voices count as "artificial" under existing robocall consent rules, which matters if you use cloned voices for outbound calling. In the EU, a voice recording is personal data under GDPR, so you need a lawful basis, usually consent, to collect and process it, and the EU AI Act adds a separate duty to tell people when audio has been artificially generated.
In practice, this means three things: don't clone a voice without documented consent from the person it belongs to, check each provider's specific terms on commercial use of cloned voices (they vary more than you'd expect), and if you operate across several countries, assume the strictest rules apply rather than the most convenient ones.
Where Voice Fits Alongside AI Video and Music
Voice rarely stands alone in a real production pipeline. Narration and dubbing pair naturally with AI-generated video, either as a voiceover on a generated clip or as a translated dub of something recorded in another language. Background music fills out a short video or ad that would otherwise be silent. A single API that covers image, video, and music generation alongside voice makes this kind of pipeline easier to build, since you aren't juggling separate vendors, API keys, and billing dashboards for what ends up as one output. If you already generate video through something like Apiframe's AI video API, adding narration or dubbing to the same pipeline takes far less work than bringing in a separate voice vendor.
You can also see how Apiframe approaches other media workflows in its text-to-video Python tutorial, which walks through authentication, sending a generation request, checking on progress, and handling the finished video.
For music, Apiframe's roundup of the Best AI Music Generation APIs in 2026 compares services such as Suno, Udio, and ElevenLabs Music on quality, pricing, and licensing.
FAQ
What is the cheapest text-to-speech API?
Amazon Polly is among the cheapest at around $4 per million characters for standard voices, and Cartesia's entry plans come close. Deepgram Aura-2 sits in the middle at $30 per million, still well under ElevenLabs' rates. Cheapest isn't always best, though, so weigh naturalness and voice cloning support against what you actually need.
Which text-to-speech API sounds most natural?
ElevenLabs consistently ranks highest for naturalness and expressiveness in independent comparisons, especially for cloned or emotional voices. For plain narration, several competitors are close enough that the difference is hard to hear outside a direct side-by-side test.
Do these APIs support real-time streaming?
Most of the major providers do to some degree, but Cartesia is built around it and publishes the lowest delay in the category. If real-time responsiveness is a hard requirement, measure it yourself rather than trusting a headline number, since real-world performance depends on your network path and the length of your text.
Is there a free text-to-speech API tier?
Most providers offer a limited free tier or trial credit for testing. ElevenLabs, PlayHT, and Cartesia all do, though none are meant for production use beyond light testing. You can also get an API key and start with free credits if you want ElevenLabs voice generation alongside video, image, and music models in one integration.