Home › Products › s2speech

Vietnamese voice AI platform

s2speech: Speech AI built for Vietnamese, ready for business

s2speech brings speech-to-text, text-to-speech and the aivision-L1.0 language model together on one platform: one account, one API key, one balance. Teams transcribe calls and meetings, generate natural voices for contact centres, and summarise and analyse Vietnamese conversations. Every account gets $10 of free usage per day, no card required.

Use it now at s2speech.com

Published figures

  • 11.84% average word error rate (WER) on 4 Vietnamese test sets
  • $10 of free usage per day for every account, no card required
  • 24 languages in the translation and meeting apps
  • 8 concurrent sessions on one ~8 GB GPU (speech-to-speech)

Lower WER is better. AIVISION internal results on data not used for training: FLEURS-vi 4.58%, VIVOS 6.83%, ViMedCSS (medical) 15.68%, real business meetings 20.29%.

Key features

  • Speech-to-text: Real-time streaming over WebSocket plus file and URL transcription over REST. Handles Vietnamese mixed with English, numbers, money and dates, with word-level timestamps and regional accents.
  • Text-to-speech: Natural Vietnamese and English voices streamed from the first word, with 8 kHz G.711 μ-law/A-law output for Asterisk and FreeSWITCH contact centres.
  • Voice cloning: Create a custom voice from 20 seconds to 2 minutes of recording, with the speaker's consent.
  • aivision-L1.0 LLM: An OpenAI-compatible API: just change the base URL. Summaries, Q&A, minutes, translation and conversation analysis, with streaming, in Vietnamese and English.
  • Ready-made apps: Hana for voice chats with an AI character, AI Voice Note for recordings and minutes, live translated captions, voice interpreting and multilingual meeting rooms; the translation and meeting apps support 24 languages.
  • On-premise speech-to-speech: The free speech-to-speech release runs in real time and deploys on-premise on a single GPU with about 8 GB of VRAM (RTX 2080 Ti class), with 8 concurrent sessions.

How it works

  1. Create an account: One account, one API key and one balance for every service, with $10 of free usage per day.
  2. Connect: Call REST or WebSocket for speech; for the LLM, keep your OpenAI library and change the base URL.
  3. Process: Transcribe, synthesise speech, summarise or translate, with results streamed back.
  4. Scale: Top up by bank transfer from 50,000 VND, or deploy speech-to-speech on-premise.

Where it fits

  • Contact centres
  • Meeting minutes
  • Live captions and translation
  • Voice interpreting
  • Conversation analytics
  • Voice notes

Deployment experience in Vietnam, the USA, Mexico, the Philippines and Thailand.

Visit s2speech.com · See all products

More from AIVISION

  • VTraks: AI retail display monitoring
  • AIV Camera: Turn existing cameras into AI cameras
  • AIV Office: AI office suite with the Hanna assistant