Speech and language AI, native in English, Thai, and Mandarin
Text to speech, transcription, OCR, and translation in one API. Paxa TTS Flash v2 speaks all three languages with natural, expressive delivery, even in the same sentence.
100 free credits to start, no card required.
Cookie
0:06Hi there, and thanks for calling. I can help you with your order, your account, or anything else.
20 / 50
Hear English, Thai, and Mandarin
Every voice you hear was generated by Paxa TTS Flash v2 from the text on screen, direction included.









Foi Thong, Narrator
Line to speakPaxa TTS Flash v2 Pixie
[whispering] [excited:2] [crying] [laughs]
The Speech page takes your own text, with word timestamps and audio you can download.
Open the Speech pagePricing
One credit balance spends across every API, at the same rates on every plan.
- Speech credits per 1,000 characters
- 10
- Transcription credits per minute of audio
- 8.33
- OCR credits per page read
- 6.5
- Translation credits per 1,000 characters translated
- 25
Plans, refills, and the market comparison live on the pricing page.
See pricingProducts
Text-to-Speech
Text to speech in English, Thai, and Mandarin from Flash, with streaming, word timestamps, and a live websocket.
Recording0:11
Transcriptth
สวัสดีค่ะคุณสมชายยอดชำระของท่านจำนวน พันสองร้อยห้าสิบ บาทครบกำหนดวันที่ สิบห้า สิงหาคมนี้ชำระผ่านแอปได้ตลอด ยี่สิบสี่ ชั่วโมงค่ะ
timestampsconventionvocabulary1.66 credits for this recording
Speech-to-TextResearch preview
Transcription for Thai recordings and other languages, with word timestamps and a spoken or written convention.
OCR and ExtractionAvailable now
Reads the documents Thai businesses actually process, from receipts and tax invoices to forms, into Markdown or structured data, or straight into the fields a schema names.
AI TranslationAvailable now
A translation API that writes native Thai, with formality, glossaries, and document context on every request.
Models & voices
From news anchor to ASMR and anime hero to grandmother, and every voice speaks English, Thai, and Mandarin.
Paxa TTS Flash v2Pixie
Our flagship text-to-speech model for natural English, Thai, and Mandarin speech.
- Languages
- EN · TH · ZH
- Voices
- 50
Pro, the next generation of Paxa speech synthesis, is coming soon.
See the text to speech APIBuild with one API
Read the docs- Streaming, two ways
- Chunked audio over plain HTTP, or a live websocket that speaks while your LLM is still writing.
- OpenAI-compatible
- The speech endpoint answers the OpenAI SDK at /v1/audio/speech. Swap the base URL and nothing else.
- Timed speech
- Ask for timestamps and speech returns word, sentence, or utterance spans for captions and read-along interfaces.
- Usage-based billing
- Speech at 10 credits per 1,000 characters, transcription at 8.33 credits per minute of audio, OCR at 6.5 credits per page, and translation at 25 credits per 1,000 characters, counted per request. Failed requests are refunded automatically.
curl -X POST https://api.paxalabs.com/v1/tts \ -H "Authorization: Bearer $PAXA_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "text": "[whispering] I can'\''t believe it. [ecstatic] We actually did it!", "voice": "foithong", "model": "paxa-tts-flash-v2" }' \ --output foithong.mp3
# Encode without line wrapping: wrapped base64 breaks the JSON string. AUDIO=$(base64 < meeting.m4a | tr -d '\n') # --max-time covers an hour-long recording; curl defaults to no limit. curl -X POST https://api.paxalabs.com/v1/stt \ --max-time 300 \ -H "Authorization: Bearer $PAXA_API_KEY" \ -H "Content-Type: application/json" \ -d "{\"audio\": \"$AUDIO\", \"model\": \"paxa-stt-lite-v1-preview\"}"
# Encode without line wrapping: wrapped base64 breaks the JSON string. DOC=$(base64 < invoice.pdf | tr -d '\n') # --max-time covers a multi-page document; curl defaults to no limit. curl -X POST https://api.paxalabs.com/v1/ocr \ --max-time 300 \ -H "Authorization: Bearer $PAXA_API_KEY" \ -H "Content-Type: application/json" \ -d "{\"document\": \"$DOC\", \"model\": \"paxa-ocr-lite-v1\"}"
# --max-time covers a large request; curl defaults to no limit, # but most clients and SDKs time out well before a big translation ends. curl -X POST https://api.paxalabs.com/v1/translate \ --max-time 300 \ -H "Authorization: Bearer $PAXA_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "text": "Welcome to our store. Free shipping on orders over 500 baht.", "model": "paxa-translation-lite-v1" }'
Timed speech
Nom Yen reading one line, lit by the word spans the timestamps option returns.
Enterprise
Realtime speech, on capacity reserved for your workload.
- Milliseconds to first audio
- 15-20
- Faster than real time
- 190-300×
For voice agents, live dubbing, and anything a person is waiting on. Volume, rate limits, and concurrency are quoted to the workload you run. Both figures are the best results from our own load test on dedicated hardware, and the self-serve plans run on shared capacity.
Use cases
- Call centers
- Voice agents
- AI companions
- E-learning
- Publishers
- Podcasts
- Audiobooks
- Audio novels
- Interactive fiction
- Creators
- Donation alerts
- Advertising
- Accessibility
- Announcements
- Call recordings
- Meetings
- Video captions
- Interviews
- Lectures
- Voice notes
- Media monitoring
- Dictation
- Live transcription
- Invoices and receipts
- Contracts
- Identity and forms
- Archives
- Knowledge bases
- Ecommerce
- Support
- Travel
- Games and apps
Built in Bangkok
A research lab in Bangkok inventing speech and language models for English, Thai, and Mandarin speakers.
- Languages, English, Thai, and Mandarin
- 3
- Models in the lab
- 4
- Voices in Paxa TTS Flash v2
- 50