Trusted AI·Local languages
DocsTrust CentreAbout Proto
Book a demo
Voice API

Local speech, your service

Plug the Proto text-to-speech and speech recognition engine into your app with prepaid credits.

Get API keysRead the docs →
Language
Speed
The API

Two endpoints

Both authenticate with a bearer token scoped to your workspace, and both draw on the same voice-credit balance.

POST /tts

Text-to-speech

Synthesis built from native recordings, not an English voice bent towards a local accent.

  • Native speaker recordings, not converted English models
  • Voice, gender and accent selectable per language, speed 0.25–4.0
  • Up to 5,000 characters a request, returned base64-encoded as MP3 or WAV
POST /asr

Speech recognition

Transcription built for regional accents and sentences that move between a local language and English mid-clause.

  • Tuned for regional accents and mixed-language phrasing
  • Transcribes up to 15MB of MP3 or WAV audio a request
  • Transcript returned with an optional English translation
Pricing

Pay as you go

Access the Voice API through our Starter plan. The rate decreases as usage grows.

credits a month
$0a month, a blended $0.000 a credit
Claim 50 credits
First 10,000$0.050/credit0$0
10,000 – 100,000$0.040/credit0$0
100,000 – 1,000,000$0.025/credit0$0
1,000,000 – 5,000,000$0.010/credit0$0

Its own balance

Voice requests draw on prepaid voice credits, not on your platform credit volume. The two balances are billed and tracked separately.

1 credit a request

Every successful text-to-speech or speech-recognition call costs 1 credit, at any tier. A failed request is not charged.

50 credits free

Every new workspace starts with 50 voice credits, no plan or purchase required to try it.

No plan required

Voice credits can be bought on any plan, from Starter plan upward — there is no tier to reach first.

What comes with every credit

Auto top-up, low-balance alerts, usage you can break down by request, and every language included from day one – not extras, just how the balance works.
Explore features & pricing
Auto top-up
Set a threshold and a further balance is applied automatically, so a live conversation never stops mid-call for want of credit.
Voice creditsBilling
Auto top-up enabled
Top up when balance falls below100credits
Then add5,000credits
Low-credit alerts
Get notified by email and in-platform before your balance runs out, well before a request starts failing.
Low voice credit balance2m ago
42 credits left – top up to keep text-to-speech and speech recognition answering.
Sent toEmailIn-platformTop up
Usage you can attribute
A dedicated usage tab breaks consumption down by request type, teamspace and language.
UsageCredits
Last 30 days
RequestLanguageCredits
TTSKinyarwanda412
ASRCebuano268
TTSTagalog151
Total831
All languages included
Every one of Proto's 92 voice languages is available on the same balance – nothing extra to unlock a language.
Voice languageAll included
Search languages
filTagalogTTS · ASR
cebCebuanoTTS · ASR
rwKinyarwandaTTS · ASR
kjOshiwamboTTS · ASR
paPunjabiTTS · ASR
urUrduTTS · ASR
+86 more languages
Local languages

Reach the last mile

Proto works at the frontier of low-resource languages.

KinyarwandaDeployed
Rwanda
Speakers
15M
ASR accuracy
93.73%
Word error rate
6.27%
ASR fallback
4.93%
ASR speed
1.75x
Voice quality (MOS)
4.00
View benchmarks →
TagalogDeployed
Philippines
Speakers
45M
ASR accuracy
90.66%
Word error rate
9.34%
ASR fallback
4.21%
ASR speed
1.74x
Voice quality (MOS)
4.00
View benchmarks →
CebuanoDeployed
Philippines
Speakers
22M
ASR accuracy
87.74%
Word error rate
12.26%
ASR fallback
6.36%
ASR speed
1.76x
Voice quality (MOS)
3.80
View benchmarks →
OshiwamboDeployed
Namibia
Speakers
1.4M
ASR accuracy
70.22%
Word error rate
29.78%
ASR fallback
8.34%
ASR speed
1.77x
Voice quality (MOS)
3.60
View benchmarks →
IlocanoIn development
Philippines
Speakers
9M
ASR accuracy
Not yet published
Word error rate
Not yet published
ASR fallback
Not yet published
ASR speed
Not yet published
Voice quality (MOS)
Not yet published
View benchmarks →
PunjabiIn development
Pakistan
Speakers
100M
ASR accuracy
51.50%
Word error rate
48.50%
ASR fallback
Not yet published
ASR speed
Not yet published
Voice quality (MOS)
Not yet published
View benchmarks →
SindhiIn development
Pakistan
Speakers
23M
ASR accuracy
Not yet published
Word error rate
Not yet published
ASR fallback
Not yet published
ASR speed
Not yet published
Voice quality (MOS)
Not yet published
View benchmarks →
UrduIn development
Pakistan
Speakers
100M
ASR accuracy
79.40%
Word error rate
20.60%
ASR fallback
Not yet published
ASR speed
Not yet published
Voice quality (MOS)
4.15
View benchmarks →

Start with 50 free credits

No plan required. Get API keys and make your first request in minutes.

Get API keysVoice API reference →︎

15‑minute live demo

Ready to connect with our team?