Rates and claims last checked August 2026
Paxa Labs and Cartesia for Thai speech
A wide multilingual speech platform and a Thai speech API compared on voices, regional registers, real-time delivery, and what a failed request costs.
Cartesia builds a low-latency speech platform across a wide language list, with Thai among the codes its API accepts. We serve two languages and nothing else. The question a buyer is actually answering is whether Thai is one locale in a product shipped everywhere, or the language the product is in.
What differs
| Dimension | Paxa Labs | Cartesia |
|---|---|---|
| What the product is shaped to do | Thai and English speech from one model, with a roster cast for Thai listeners. | A speech platform across 47 language codes, Thai among them, on the sonic model line. |
| Thai voices | 26 in the roster, each with a published id, a fixed sample line, and a clip anyone can play without an account. | Their voice library is where to check the Thai entries. |
| Regional Thai registers | Isan, Northern, and Southern voices, each with a published clip and a page of its own. | Their voice library is where to check what registers they carry. |
| Real-time delivery | Chunked HTTP streaming and a live websocket. Audio starts before the full text exists. | A websocket endpoint for realtime synthesis, documented in their API reference. |
| Word timestamps | Word, sentence, or utterance spans on the same call, at a stated multiplier on the per-character rate. | An add_timestamps parameter that returns word-level timing with the audio. |
| Voice cloning | None. The roster is fixed and every voice keeps a stable public id across model generations. | Voice cloning is part of their platform. |
| What a failed request costs | Nothing. Credits are charged before inference and refunded automatically when generation fails, including mid-stream. | Their billing terms are on their pricing page. |
- Paxa TTS Flash, per 1M characters
- $15
- Cartesia
- $50
- Their rate, as a multiple
- 3.3×
Both figures are published list rates normalized to one million characters, read from each vendor's own page and re-checked on a fixed schedule.
What it costs at your volume
| Speech, per month | Paxa Labs | Cartesia | Difference |
|---|---|---|---|
| 100,000About 1.8 hours of speech | $1.50 | $5 | $3.50 less |
| 1,000,000About 18 hours | $15 | $50 | $35 less |
| 5,000,000About 92 hours | $75 | $250 | $175 less |
| 20,000,000About 370 hours | $300 | $1,000 | $700 less |
Every figure is the two published rates multiplied by the volume beside it. Where the other vendor publishes a range, the low end is compared and the range is shown.
When Cartesia is the better choice
If your product speaks in a dozen languages and Thai is one of them, one vendor across all of them beats a better voice in a single locale. Cartesia lists 47 language codes and we list two. Cloning is theirs as well, and we have none, because a fixed roster with stable public ids is the trade we chose. Choose us when the Thai read is the part your users judge you on.
Questions
- Does Cartesia support Thai?
- Their API reference lists th among its language codes. What their Thai voices sound like is a question for their voice library and your own ears.
- Do you offer voice cloning?
- No. Every voice is cast, published with a stable id, and auditionable before you write a line of code. A roster that cannot be extended on demand is the cost of that.
- Do you publish quality comparisons?
- No. We publish no quality, accuracy, or latency figures of our own, for our model or anyone else's. Generate the same Thai script on both and listen.
- How current are the prices here?
- They come from each vendor's published rates, recorded in one place and re-checked on a fixed schedule; the month of the last check is at the top of the page.
Start with free credits
Sign in with Google, GitHub, or Hugging Face and spend 100 one-time free credits on your own text.