The surfaces nobody transcribes
Everything a business says aloud goes almost entirely unaudited, for a straightforward reason. Reviewing it means listening to it, and listening does not scale the way scanning a page does. So audio and video become the place where abandoned claims survive longest, where three different tones coexist happily, and where a brand can contradict itself for years without anybody noticing.
Two kinds of voice
The literal voice: who speaks for the business, in what accent and register, at what pace. Frequently a decision nobody made, arrived at through whoever was available on the day. And the written voice, read aloud. A brand whose copy reads considered while its narration sounds like a hard-sell radio spot has two identities. The audible one usually wins, since it reaches people who never read the page.
Sound as identity
Music, stings and any recurring sonic element. Sonic identity is a genuine asset and almost never governed, so most businesses have as many musical characters as they have had editors. Pacing and edit rhythm carry character exactly the way motion easing does on a screen. Fast and hard reads as urgent, while slow and held reads as considered. Applied inconsistently across a library, it reads as unfinished.
Every spoken claim, transcribed and checked
The highest-value part of this pass. Assertions made aloud are rarely written down and therefore rarely checked, which is precisely why lapsed certifications, abandoned superlatives and superseded guarantees survive longest in recorded material. Everything gets transcribed and every factual claim gets the same primary-source verification described in claim verification. Certifications against the issuing register. Founding dates against the record. Superlatives against whatever evidence exists.
The silent version is the real version
Most video is watched muted, at least at first. If the message depends on audio, it is not landing. We check what each piece communicates with sound off and no captions, which is how a large share of its audience actually encounters it. Captions matter twice over. They are an accessibility requirement in many contexts, and they are the only machine-readable text in the entire asset. Auto-generated captions routinely mangle product names and technical terms, which then propagate wherever the transcript travels.
The derivative cuts
The short version, the vertical crop, the platform-specific edit. Then the thumbnail, the end card and the lower thirds. These are usually produced by whoever was editing, at speed, outside any system, though they are what most people actually see.
The phone
For many businesses the first human contact, and governed by nobody. The greeting, the menu, the hold music, the voicemail, and the tone of whoever answers on a difficult afternoon. None of it appears in any brand book we have examined.
Borrowed platforms
Podcast appearances, interviews, panels and webinars where somebody spoke for the business on a property it does not control. Inventory what exists, what was claimed, and who can correct it if anything was wrong. Material sitting on an agency account or a former employee’s channel is not yours to fix.
Longevity
Audio outlives video, and video outlives almost everything. Episodes stay in feeds indefinitely. Old material stays reachable for years, still making claims the business abandoned, still showing staff who left, still quoting prices and locations that changed. The inventory covers what a member of the public can still find rather than what marketing is currently running.
Synthetic and cloned voice
Our production standard is that no AI-generated or cloned voice ships without written consent from the person whose voice it derives from, and a partially modified voice counts as cloned. Where an existing asset uses one and consent cannot be evidenced, that is recorded as a fact and handed to you to resolve.
Where voice work connects
Written claims are covered in claim verification. How the same identity looks rather than sounds is covered in visual continuity. Motion timing as brand character is covered in motion design under interactive experience design.
How the engagement runs
Screen shared sessions, with transcription done rather than skipped, because the claims nobody has checked are the point of the exercise.
"If nobody has listened to your hold music since it was set up, somebody hears it every day."