AssemblyAI Speech-to-Text and Voice Agent API

AssemblyAI, Inc. is a voice AI company founded in 2017 and headquartered in New York City, led by CEO Dylan Fox. It has raised 130 million USD from investors including Accel, Insight Partners, Smith Point and Y Combinator, and reports more than 120 million conversations handled per week on its API.

Its speech models convert audio to text through one API for both batch files and real-time streams. Universal-3.5 Pro adds native code-switching and a medical mode, Universal-2 covers 99 languages for high-volume work, and U-3.6 Pro targets real-time voice agents and AI scribes, all with Vietnamese support. Add-ons include speaker diarization, entity detection, PII redaction, keyterm prompting, translation and subtitle export, while an LLM Gateway and a Voice Agent API let developers build complete voice bots. Customers are software companies building meeting notes, call analytics, contact center and voice agent products.

• Universal-3.5 Pro: async model with native code-switching, improved diarization and medical mode.
• Real-time streaming: U-3.6 Pro for voice agents and AI scribes with low latency.
• Speech understanding: diarization, entity detection, language detection and PII redaction.
• Keyterm prompting: boosts recognition of names, products and domain terms.
• Voice Agent API and LLM Gateway: builds voice bots and applies LLMs to transcripts.
• Output formats: translation, multi-channel transcription and SRT or VTT subtitles.

Vietnamese software startups can add Vietnamese transcription and summaries to CRM, meeting or call center products through the API. Contact centers can analyse recorded calls in Vietnamese and English with PII redaction. Developers can build real-time Vietnamese voice agents using the streaming model.

Vietnamese is priced the same as other languages, from 0.15 USD per hour on Universal-2 to 0.45 USD per hour for real-time voice agent use, with unlimited concurrency on all listed models. One API handles both live and recorded audio, so teams do not need separate vendors. PII redaction and a HIPAA BAA on request help with sensitive call data.

Voice agents, AI meeting assistants and call analytics are among the fastest growing AI product categories, and each needs accurate speech recognition. Builders in Vietnam and Southeast Asia need models that handle local languages and mixed-language speech at low latency and low cost per hour.

AssemblyAI, Inc. was founded in 2017, is headquartered in New York City and is led by CEO Dylan Fox. Total funding is 130 million USD at Series C from Accel, Insight Partners, Smith Point and Y Combinator. Customers include Zoom, Granola, HeyGen, ClickUp, Apollo.io, CallRail and Retell.