About
DUSKOsound builds the entire audio identity of a project: sound design, original music, foley, and final mix for films, video games, and animation. One author, one coherent sound.
Working notes on audio in AI-assisted production, for linear and interactive media.
For the past year I’ve been working inside AI-assisted production pipelines, in both linear media (film) and interactive media (games), with one focus: understanding where audio best contributes to AI integrated media production pipelines. The pattern is consistent: picture arrives ahead of sound, so audio is the layer that determines whether the result reads as a film.
• Prompt “no music, natural sound only.” Every clip arrives with clean, usable production sound instead of baked-in audio you have to fight in the mix. (Kavan The Kid, “The Chronicles of Bone”, Magnific BTS session)
• Deliver dialogue for lip-sync as a black-screen MP4. Models weight video reference over audio reference; the black frame avoids lip-graft artifacts. (same source; part of a delivery spec I keep refining)
• When generated lip-sync fails, post-sync the dialogue. Classic ADR beats re-rolling shots, cheaper, faster, better.
• Loudness-normalize every generated SFX. AI sound tools swing wildly in level (I’ve measured ±30 dB). Nothing enters the session without normalization.
• Cut to a beat sheet. When the edit is built on beats, score and sound design hit them by design, not luck. (PJ Ace / Genre AI, Wonder Studios workflow session)
• Define the sonic identity first. Decide what the world sounds like at rest and every scene inherits coherence, the audio analog of light-first cinematography. (cf. Waviboy, “Fifty Frames, Decoded”)
• Build the physical world by hand. Generated footage has no room tone, no footsteps, no space. On “Black Harvest” (dir. Giulio Fulchignoni) we authored the entire sonic reality from zero.
• Realism lives in post. Generated picture and generated sound both earn their believability in the finishing pass, not the prompt.
On the game side I’m producing an AI-assisted title end to end. There, audio’s contribution extends past assets into implementation: state machines, adaptive music systems, engine integration.
Sound & music for AI films
Book A Call
DUSKOsound builds the entire audio identity of a project: sound design, original music, foley, and final mix for films, video games, and animation. One author, one coherent sound.