API Sunsetting

Over the past six months, we have worked closely with teams using both the ai-coustics SDK and the ai-coustics API. During this time, customer demand has become clearly concentrated around real-time speech enhancement for Voice AI systems. To fully focus on this segment, we will discontinue the ai-coustics API on March 31, 2026. As a result, our Lark and Finch models will no longer be available for use. Our long-term focus is on real-time Voice AI pipelines and on-device SDK deployments, including VAD, Voice Focus, and ASR-optimized preprocessing. Maintaining a general-purpose API limits our ability to invest fully in these areas.

​

Unused credits

Any remaining prepaid balances will be refunded or, if preferred, converted into SDK credit. We will reach out individually with your remaining balance details.

​

Continuing with ai-coustics

Our real-time SDK remains fully supported for teams building Voice AI applications, offering low-latency processing, on-device or on-premise deployment, built-in VAD, and Voice Focus. We appreciate every team that has built with the ai-coustics API. This step allows us to focus fully on reliable Voice AI infrastructure.

​

SDK vs. API

The ai-coustics API processed audio files. That means you uploaded a recording and received an enhanced version back. The SDK works differently: it processes audio in real time, frame by frame, as a stream. It is designed to sit inside a live audio pipeline (e.g. a voice call or a Voice AI agent), not to batch-process files. An SDK playground is available in our Developer Platform, and a Hugging Face demo allows you to try Voice Focus without an account. However, these playgrounds do not support downloading enhanced audio files, as the SDK is not a product designed for the enhancement of pre-recorded audio for content creation.

​

Contact