Paxa Labs and Soniox for Thai speech to text

A transcription API that is Thai first, next to Soniox's file and realtime models, across language coverage, Thai mixed with English, request flow, speakers, domain terms, timestamps and subtitles, and realtime turns.

Soniox sells Thai transcription for recorded files and live streams, at one published rate for every language. Paxa STT Lite transcribes Thai, English, Mandarin, and other languages, with word timestamps, speaker turns, a spoken or written convention for numbers, and subtitles on the same request. It is a Research preview, and its transcript quality and options may change between releases.

Sources:Soniox Thai speech to text(opens in a new tab)

Rates and claims last checked October 2026

Start free See full pricing

What differs

What differs
Dimension	Paxa Labs	Soniox(opens in a new tab)
Languages	Thai, English, Mandarin, and other languages. Paxa STT Lite writes speech in the language spoken and detects that language itself. A BCP 47 hint corrects a recording it misreads.	60+ languages, Thai among them. A language hint biases recognition, and language identification adds a language field to every token.
Thai and English in one recording	A Thai sentence that switches into English is written as spoken, with no setting. Nothing is translated.	Their Thai page says the model follows switches between Thai and other languages, within the same sentence too.
Request flow	One request returns the transcript in its response, for recordings up to 60 minutes. Split a longer recording at a pause.	A local file is uploaded to /v1/files, a POST to /v1/transcriptions starts the job, and you poll or take a webhook, then fetch the transcript from its own endpoint. A file runs up to 300 minutes.
Speakers	Speaker turns numbered from 0 in order of first appearance, on recordings under 9 minutes, at no extra charge.	A speaker field on every token, up to 15 speakers per session, included in the hourly rate. Their guide recommends file transcription for the most accurate separation.
Domain terms	A vocabulary of up to 50 terms, in Thai or any other language, at no extra charge.	A context object takes key-value details, free text, a terms list, and translation terms, up to 8,000 tokens (about 10,000 characters), billed by token usage.
Timestamps and subtitles	Word timestamps with Thai segmented into words, and an SRT or WebVTT file in the same response. Every line break falls on a word boundary, Thai included.	A start_ms and an end_ms on every token, and a token can be part of a word.
Realtime turns	Paxa STT Lite Realtime bills the audio it receives, silence included, and reports when the user starts and stops talking.	Endpoint detection returns an <end> token when the model judges the speaker finished, weighing pauses, intonation, and context. max_endpoint_delay_ms caps the wait between 500 and 3,000 ms.

Sources:Soniox pricing(opens in a new tab)Soniox Thai speech to text(opens in a new tab)Soniox async transcription(opens in a new tab)Soniox async limits(opens in a new tab)Soniox speaker diarization(opens in a new tab)Soniox context(opens in a new tab)Soniox timestamps(opens in a new tab)Soniox endpoint detection(opens in a new tab)

Paxa STT Lite, per hour of audio: $0.50
Soniox: $0.10
Their rate, as a multiple: 0.2×

Their rate is the file transcription rate, the same for every language and billed by token. Realtime streaming has its own published rate. Every rate is per hour of recorded audio, from the vendor's cheapest pay-as-you-go model to its most expensive. Per-minute and per-second rates are converted to that unit.

What it costs at your volume

What it costs at your volume
Audio, per month	Paxa Labs	Soniox	Difference
10 hours	Paxa Labs$5	Soniox$1	Difference$4 more
100 hours	Paxa Labs$50	Soniox$10	Difference$40 more
1,000 hours	Paxa Labs$500	Soniox$100	Difference$400 more
10,000 hours	Paxa Labs$5,000	Soniox$1,000	Difference$4,000 more

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 Soniox is the better choice

Their published rate for recorded audio is below ours, and a file runs up to 300 minutes where ours stops at 60. Their realtime API also translates as people speak, and processes in the US, EU, Japan, or India. Ours is a Research preview, and its output and options may change between releases. Choose us when you want the transcript, word timestamps, speaker turns, and Thai subtitles back from one request, with numbers written the way your text needs them.

Questions

Does Soniox transcribe Thai?

Yes. Their Thai page offers Thai transcription in real time over WebSocket and from recorded files, at the same rate as every other language.

What does Research preview mean for Paxa STT Lite?

Requests bill at the published rate. Transcript quality and the option set may change between releases, and the changelog on our speech to text page records each change.

Do you publish accuracy or latency comparisons?

No. We publish no accuracy, quality, or latency figures of our own, for our model or anyone else's. Transcribe the same recording on both and read the two transcripts.

How current are the prices here?

They come from each vendor's published list rate, recorded in one place and re-checked on a fixed schedule. The month of the last check is at the top of the page.

Other comparisons

Start with free credits

Start free