Audio Denoise & Repair
Choose speech denoise, music repair that keeps vocals and instruments, or advanced target sound separation. Every AI model runs only on the self-hosted remote GPU.
FAQ
It is designed for meetings, interviews, podcasts and spoken recordings. MossFormer2 keeps speech while reducing hiss, fan noise, electrical hum and some complex background noise, with 48kHz mono output.
No. Music repair keeps vocals, drums, bass and other instruments, applies gentle cleanup to each part and remixes the full song.
It can keep or remove a described sound such as applause, wind, traffic, speech or music. This advanced mode outputs 32kHz mono and is not intended for high-fidelity music repair.
Yes. All modes upload audio to the self-hosted GPU service. No AI model runs on your computer.