Tool profile
Soniox: cost, fit and limits
Soniox is a speech AI vendor that sells an API for transcription and real-time translation priced by audio hour, with text to speech on the same price list. The pricing page names its current speech to text model as STT v6.
What to know first
Soniox prices by tokens but prints hourly equivalents: $0.10 per hour for async file transcription and $0.12 per hour for real-time streaming. Diarization, language identification and smart formatting are said to be bundled into the real-time rate. There is no subscription and the pricing page mentions no free tier. The legal entity, country and data practices are not stated on the page we read.
Soniox at a glance
| Maker | Soniox (legal entity not stated on the opened pages), Not stated on the vendor pages |
|---|---|
| Free plan | No |
| Entry price | $0.10 per hour |
| API access | Yes |
| Category | Best AI Speech to Text APIs in 2026, rank 2 of 12 |
| Website | soniox.com |
Who Soniox is best for
High-volume batch transcription
At $0.10 per hour for files, backlogs of recorded calls or lectures cost less per hour than any other printed rate in this category.
Live products needing speaker labels
Diarization and language identification are listed as part of the $0.12 real-time rate, so those features do not appear as separate line items.
Teams building voice output as well
Real-time text to speech is sold on the same price page at $0.70 per hour, which keeps one vendor and one invoice for both directions.
Soniox plans and pricing
| Plan | Price | What it includes |
|---|---|---|
| Speech to text, async | $0.10 per hour | File transcription at $1.50 per 1M audio input tokens. Translation or other text output is billed as output tokens. |
| Speech to text, real-time | $0.12 per hour | Streaming transcription at $2.00 per 1M audio input tokens, with diarization, language identification and smart formatting said to be bundled. |
| Text to speech, real-time | $0.70 per hour | Generated speech from text, at $4.00 per 1M input text tokens and $21.50 per 1M output audio tokens. |
| Enterprise | Custom package | Enterprise discounts and committed-use contracts may differ from the listed rates. The page does not say how large a commitment qualifies for a discount. |
Checked November 2026.
What Soniox does
- Real-time streaming
- Streaming transcription is billed per hour of audio and bundles diarization, language identification and smart formatting into the single real-time rate. That bundling means a speaker-labeled live feed carries no separate surcharge on the printed price list.
- Async file mode
- File transcription costs less than streaming and can add translation, with translated text charged as output tokens on top of the audio input. Output tokens for translation come on top, which matters for multilingual workflows more than for plain English transcripts.
- Token billing with hourly equivalents
- The page prices audio in tokens, $1.50 per million for files and $2.00 for streaming, then restates each as an hourly figure for easier budgeting.
- Text to speech
- Speech synthesis is sold on the same page, which suits products that listen and reply, though it is outside the scope of this ranking. The page presents the two products side by side, though the synthesis rate sits far above the transcription rates.
How Soniox handles your data
The Soniox pricing page does not state whether customer audio is used for training, how long it is retained, where it is processed or which certifications the company holds. A Policies page is linked from the pricing page but was not opened for this entry, so no claim is made about it. Because the legal entity and country are also unstated on the page we read, a buyer subject to GDPR should request the data processing agreement and a list of subprocessors first.
Where Soniox falls short in practice
- No free tier mentioned. The pricing page lists no trial credit or monthly free allowance, so testing means paying real money from the first request, even if the hourly rate is low.
- Token-based billing. Costs are metered in tokens, and the hourly figures are the page's own equivalents. Your language mix and speech density may produce a different bill.
- Enterprise terms hidden. Committed-use and discount terms are described as possibly different, with no figures, so large buyers cannot model contract pricing in advance. Without a visible free tier, the first real test of accuracy on your audio carries a cost, however small.
Works with
Alternatives to Soniox
Soniox: common questions
How much does Soniox cost per hour?
The pricing page lists $0.10 per hour for async file transcription and $0.12 per hour for real-time streaming. Those are hourly equivalents of token prices, $1.50 and $2.00 per million audio input tokens. Translation or other text output is billed separately as output tokens, so a translated workflow costs more than the headline rate.
Does Soniox have a free plan?
The pricing page does not mention a free tier or a trial credit. It describes pay as you go API pricing with no subscription plan listed. If a free allowance exists, it was not on the page we read, so ask the vendor or check the dashboard after signing up.
What is bundled into the Soniox real-time rate?
Diarization, language identification and smart formatting are said to be bundled into the $0.12 per hour real-time rate. The page lists them as part of streaming transcription rather than as add-ons, which differs from vendors that bill speaker labels as a separate per-hour feature.
Does Soniox sell text to speech too?
Yes. Real-time text to speech is priced at $0.70 per hour on the same page, built from $4.00 per million input text tokens and $21.50 per million output audio tokens. That is a separate product from speech to text and is not covered by the ranking here.
Where is Soniox based and who operates it?
The pricing page we read states neither the legal entity nor the country, so this site records both as not stated. A buyer who needs a contracting party for a data processing agreement should find the company details in the terms or ask the vendor directly.