customers
All customer stories
Top Voice AI companies are building with Assembly.
resources
Latest Release
Text-to-Speech API
Our first text-to-speech model, AssemblyAI-TTS-1, is built for voice agents, on live calls. It sounds natural, responds fast, and keeps callers on the line.
resources
The browser's SpeechRecognition is fine for a demo, but it's Chromium-biased, inconsistent across browsers, and gives you no server control. AssemblyAI's streaming speech-to-text gives you production-grade, realtime Speech-to-Text you actually control.
Let's pull up the Q3 revenue numbers before the call.
Can you transcribe this in real time across every browser?
Yes—same WebSocket, same accuracy, Chrome, Safari, or Firefox.
One WebSocket pipeline transcribes identically in Chrome, Safari, and Firefox—no Chromium lock-in.
Pick your model and where audio is processed instead of handing it to an opaque browser-vendor service with no data agreement.
Universal-3.6 Pro Realtime ($0.45/hr) streams 32 languages with speaker labels and consistent accuracy on noisy, accented, and technical speech; Universal Streaming covers English at $0.15/hr.
36%
improvement in close rate
“The new Universal-3.5 Pro speech model from AssemblyAI is best so far in terms of accuracy, latency, and language switching.”
80%
increase in customer satisfaction
“Assembly has saved us countless hours managing models, and provided exceptional accuracy.”
36%
improvement in close rate
“The new Universal-3.5 Pro speech model from AssemblyAI is best so far in terms of accuracy, latency, and language switching.”
80%
increase in customer satisfaction
“Assembly has saved us countless hours managing models, and provided exceptional accuracy.”
Capture mic audio the same way you do today, but stream it to AssemblyAI over a WebSocket instead of the browser engine. You get real-time turns that work in every browser—and you choose the model and where the audio is processed.
Start building freeNo credit card required
The browser's recognizer is a black box. Streaming word error rate measures real-time accuracy against a human reference, so you can see exactly what you gain by moving off it.
Normalized WER on the voice-agent caller corpus, mean of three noise conditions.
*Lower is better*Source: AssemblyAI realtime benchmarks technical report v2026.09 (assemblyai.com/benchmarks), data 18 Aug–24 Sep 2026.
