Welcome! This interactive demo lets you denoise and enhance speech using ai-coustics models. Choose from three options, each designed for different use cases:

• Quail: Our lightest model, created for real-time voice isolation. Try it for free or integrate via our SDK.
• Finch: A larger, subtractive model focused on voice isolation and noise removal while preserving the original voice.
• Lark: Our most advanced, generative model for universal speech enhancement. Lark not only removes noise but also reconstructs and restores lost details in audio, keeping the speaker’s voice intact.

Finch and Lark are available via our API. To get started, request an API key. You’ll receive 120 minutes for free. For additional usage, see our pricing details.

How to use
Upload or record a noisy audio sample, optionally add background noise, and adjust the enhancement level to your preference. Select the model you want to use. Note that Quail is typically the fastest option. You can also try the example clips below to explore to preview the capabilities of each model.