AI in Audio 2026: What's Hype and What Already Works
28 June 2026 · AI · realtime · trends
Generative SFX, spatial upmixing, realtime adaptation to space. A sober overview of AI tools for sound people — from practice, not marketing.
AI in audio is full of big promises. This is my sober overview of what I actually use in 2026, what I'm watching — and where I still see only demo videos.
What already works
Speech cleanup. AI denoising and dialogue repair is today's most useful AI tool in post. It can rescue material that used to be unusable. Just beware of over-processing — a "cleaned" voice with no room sounds dead.
Transcription and diarisation. Whisper and friends transcribe surprisingly well, with timing and speaker separation. For podcast editing, a text view of the conversation is a huge accelerator — you edit by transcript, not by waveform. The AI grading in my Academy is built on exactly this principle.
Generative SFX. Text-prompted sound effects work for prototyping: you quickly verify what type of sound a scene needs. The final I still almost always build from real recordings — generated sounds lack detail and variation.
What I'm watching most closely
Realtime adaptation of sound to space. Systems that change the mix in real time based on hall acoustics, visitor numbers or time of day. For galleries and museums it's the next evolutionary step: an artwork that breathes with the space. This is the subject of my research and consulting.
Spatial upmixing. AI conversion of stereo into multichannel formats is improving fast. For archive material in spatial installations it may become an essential tool.
Personalised sound zones. Directional speakers and AI-driven beamforming — two visitors a metre apart hear different things, no headphones.
Where I'm sceptical
Fully generated soundtracks and "one-button AI mixing". Sound dramaturgy is decision-making about meaning — and no model can be delegated that yet. AI is a great assistant and a poor author.
Want to feel adaptive audio yourself? The Sound & AI Lab section of this site runs an interactive demo — the same sound, a changing space, right in your browser.