Understanding AI in Audio Post: What It Means and How It’s Used

At Smart Post, AI in audio post doesn't mean just one thing.

This guide breaks down the four families we actually use – and their definitions as it pertains to audio post – ML on audio, Speech AI (ASR/TTS), LLM text and automation tools, and generative audio tools – and where they fit into our workflows (turnover, prep, cleanup, editorial, design, mixing, metadata, and delivery).

In essence, these tools handle the repeatable tasks so we can make creative decisions more fluidly.

Our "AI" motto has become: Tools speed the chores; people shape the story.

What AI Means to Us

In our practice – and in our view – when we say “AI” in audio post, what we refer to can be broken down to four practical families of tools:

• Machine learning (ML) on audio – learns from examples to clean, separate, and/or enhance sound or recorded material (think denoising, de-reverb, source separation, basic editorial/conforming, etc.).

• Speech AI (ASR/TTS) – ASR turns speech into text; TTS synthesizes voices (used only within union-compliant consent of course!) and converts "text-to-speech."

• Large-language Models (LLMs) for text/formatting automating workflows – summarizes, outlines, tags, and drafts notes and metadata; everything we do here is human-reviewed.

• Generative audio/design – references ambiences/textures/sound libraries for concepts or idea generation; not a substitute for a real/human performance, design, or mix decision – EVER.

  
  

Where AI Helps Us (Use Cases)

Turnover and Prep

• Triage problem clips (noisy, clipping, roomy mics, etc.). This helps us troubleshoot issues before we get into editorial.

• Batch loudness leveling, processing, and standardizing of any/all filenames and/or metadata tagging.

• Automate cue-sheet helpers and metadata consistency scans.

  
  

Dialogue Editorial

• Denoise/de-reverb for intelligibility with minimal artifacts.

• Source separation to rescue lines under possible music/FX and/or noise.

• Detect any "alternate" dialogue takes that may be in dailies, but not in AAF/OMF exports.

• Remove fillers and long silences in dialogue-only content.

  
  

ADR and Continuity

• Align ADR timing/pitch to production/recorded audio.

• Align boom/lav for phase-coherent compositing.

• When cleared: micro-fix line reads with approved voice modeling or pitch/timing adjustments that may be needed.

  
  

Pre-Dub and Mix

• First-pass suggestions (dialogue leveling, EQ starting points/matching).

• Cleanup/balancing that speeds up mix decisions without changing creative intent.

• Adding semantic markers for tricky words/off-mic moments that may need additional attention from the mixers.

• Object placement, upmix, and previewing formats efficiently before a final mix – especially useful in spatial audio formats.

  
  

Delivery and QC

• Run/process loudness compliance for multiple studios/networks at scale.

• Generate captions/transcripts with spot checks.

  
  

Archive and Asset Management

• Tag speakers/notes/scenes for faster retrieval.

• Keep transcripts searchable and tied to original scripts or assets associated with sound files or recorded material.

  
  

Remote Recording, Playbacks, and Review

• Low-latency conferencing and recording solutions. Playbacks/notes/fixes and hi-res client streaming for faster fixes and approvals.

  
  

The Tools We Use (...and Why)

Cleanup and Restoration (these are our current ML workhorses)

• iZotope RX – Dialogue Isolate, Repair Assistant, Spectral Recovery; daily restoration backbone.

• Steinberg SpectraLayers Pro – visual spectral editing/separation that complements RX.

• Adobe Enhanced Speech – fast web cleanup for roughs/remote recordings.

• Waves Clarity Vx Pro / Hush – real-time dialogue-focused denoise for speed passes.

• Acon Extract:Dialogue / Accentize dxRevive / Supertone Clear – alternate flavors for specific noise/artifact profiling and cleanup.

• Absentia DX – batch de-click/de-hum tailored for production dialog.

• CEDAR / VoicEX 2 – adaptive dialogue suppression when transparency is critical.

• Auphonic – intelligent leveling and LUFS compliance for voice-heavy content.

• Krisp (real-time) – keeps remote VO/ADR/Podcast audio cleaner from the streaming source.

• Zynaptiq Unveil / Unchirp, CrumplePop and Acon Digital DeVerberate for "special-use" cases.

  
  

Alignment and Continuity

• Sound Radix Auto-Align Post – dynamic boom/lav mic alignment and spectral phase correction.

• Synchro Arts Revoice Pro / Vocalign – ADR timing/pitch alignment and doubling.

• DaVinci Resolve Fairlight Voice Isolation and Dialogue Leveler – light-lift tools inside picture editorial (when needed).

  
  

Conform and Change Management

• The Cargo Cult Matchbox – compare EDLs, find differences, build reconform maps and auto-conform sessions to new picture turnovers.

• Sounds In Sync EdiLoad – merge EDLs, create conform lists, spotting.

  
  

ADR, Dubbing, ASR/TTS/Voice Cloning and Recording

• Sounds In Sync EdiCue / EdiPrompt – cue sheets and on-screen overlay of streamers/beeps/prompting.

• Non-Lethal Applications Cue Pro – ADR cueing app with on-screen streamers and live, shareable cue sheets.

• VoiceQ – dubbing/ADR with clear syllable timing overlays.

• Resemble.ai – create, edit, emote, and localize voices (either web or API). Good for temp ADR, pickups, and localization - when cleared.

• ElevenLabs – platform of AI voice tools – high-quality text-to-speech, speech-to-speech, consent-based voice cloning, multilingual dubbing, and huge voice library – handy for temp ADR and quick pickups when cleared.

• Google TTS/ Custom Voice – wide language coverage; custom models with great review process.

  
  

Remote Record and Review

• Source-Connect – industry-standard low-latency remote record and playback.

• SessionLinkPRO / Cleanfeed – browser-based talent/producer connections.

• Audiomovers ListenTo – stream your DAW bus to clients in hi-res for approvals.

• ClearView Flex – secure, low-latency streaming for remote review.

• Streambox – Spectra (software) and Chroma (hardware) – deliver secure, low-latency, streaming for remote review/collaboration.

  
  

Sound Design and Foley

• Krotos Studio / Weaponiser / Dehumaniser – performance-driven, layered SFX with quick variation; great for fast, creative ideas and design.

• Krotos Reformer Pro – “perform” Foley/textures from mic or track input (cloth, footsteps, creatures, etc.) for natural, organic ideas.

• Soundly – our hub for fast search, collections, and consistent SFX tagging/versioning, plus Voice Designer for PA/airport announcements, background conversation, and quick utility VO.

• Accentize Chameleon – AI reverb-matching plugin that analyzes a recording’s room acoustics and builds a reverb profile you can apply to ADR, Foley, or dry tracks.

• ElevenLabs Sound Effects – generates sound effects from simple text prompts, with quick variations. Used as tools, not inventory – used in context, not resold, and we stick to the provider’s terms.

  
  

Library and Metadata

• Soundly (also listed above), Soundminer and BaseHead for deep metadata tagging/scanning, DAW spotting, and alternative search workflows of sound libraries.

  
  

Loudness, Metering and Delivery

• NUGEN VisLM / LM-Correct / ISL – realtime + one-click compliance and true-peak limiting.

• iZotope Insight 2 – complete metering (loudness, surround, spectrum).

• Dolby Atmos Production Suite / Renderer – ADM, binaural renders, deliverable checks.

  
  

Spatial and Immersive

• NUGEN Halo Upmix / Downmix – transparent format moves (stereo↔5.1↔Atmos beds).

• Sennheiser Dear Reality – binaural/immersive positioning for previews and temp 3D scenes.

• Flux/IRCAM SPAT Revolution – advanced immersive mixing and room modeling.

  
  

Union Compliance (SAG-AFTRA): Digital Replicas and Voice Synthesis

Most of the actors we record are SAG-AFTRA members, so any use of a digital voice replica is governed by union rules and contracts. In practice, it's all about:

• Consent first – informed, written consent before any creation or use of a digital replica. We are completely transparent and pride ourselves on that.

• Defined scope – where, how long, and for what the replica may be used; new uses require new approvals.

• Compensation and credit – union-compliant terms per the applicable agreement.

• Security – private handling, least-privilege access, documented chain of custody.

  
  

Our policy: We only engage TTS/voice-cloning workflows when a project has documented SAG-AFTRA compliance – in accordance with the SAG-AFTRA SOUND RECORDINGS CODE (consent on file, scope defined, compensation arranged). If not cleared, we strongly urge recording traditional ADR.

The Human Element: Why Craft Still Wins

Give ten pros the same tools and you’ll get ten different results... and that’s the point!

• Taste and intent – deciding what to fix vs what to keep is storytelling, not algorithmic decision making.

• Context and judging trade-offs – sometimes a little noise, air, or a slight rustle in the recording is the "life" that would otherwise be a lifeless, artifact-free line. This will ALWAYS feel more natural to the human ear.

• Performance direction – ADR/Dialect coaching/direction, mic choices, and/or room setups change outcomes before any tool runs a "process" to dictate a performance.

• Session architecture – how editors, assistants, and techs name, route, group, and setup templates for each session helps speed up creative decision-making and prevent errors.

• Problem framing – knowing why a clip might sound wrong guides the right decision making – and avoids any additional over-processing or time wasted.

• Ears and trust – feedback discussions with directors/producers/talent, fast A/B comparisons, and delivery confidence can’t be automated.

  
  

AI can help accelerate some tasks, but it can’t model your taste, your judgment, or your collaboration and vision with a director or producer.

THIS human touch is the difference between “fixed” and finished.

What AI Won’t Do Here

• Replace actors, mixers, or editors – taste and storytelling are human.

• Decide creative intent – it can suggest; it doesn’t direct.

• Break trust – no unapproved cloning of any kind; no shady data usage whatsoever.

  
  

Results You Can Expect

• Clean dialogue tracks built faster, with fewer artifacts.

• Quicker first passes that keep creative momentum flowing.

• Consistent loudness/metadata tagging/captioning across multiple versions.

• Better exchanges between editorial, recording, mix, and delivery departments.

  
  

FAQ

Let’s Post Smart.

Ready to hear how any of these workflows could help your next project?

Request a Quote or Contact the Team and we’ll review a sample and outline options for you.