Audio scripts and voice production for streaming radio

A 30-second streaming audio spot is 72 to 78 words at the pace an actual person speaks. That single constraint decides everything after it: one idea, one buyer problem, the brand name inside the first five seconds and again in the last eight. There is no click and usually no screen in these listening sessions, so the only thing a listener can carry out of the spot is a name they can spell into a search box. How a script gets made here: We start from the search query rather than the brief. If people type a problem into Google before they type your brand, the spot names that problem in its opening line. A brief listing five product benefits comes back with one of them in it, and a note explaining which one and why the other four cost more than they returned. Two scripts minimum per campaign, written against different opening angles, because the first four seconds set the completion rate and no amount of media planning recovers a weak open. Both get read aloud at a table before anyone books a studio. Copy that survives on the page and dies in the mouth is cut there, where cutting it is free. Casting and rights: Three to five voices auditioned against your actual script, never off a showreel. Session fee plus a defined usage window, UK reads under the Equity audio agreement, US reads under SAG-AFTRA where you are a signatory, and a buyout with territory, media and term written down in one place. We have inherited too many campaigns paused mid-flight because a twelve-month buyout expired and nobody had diarised it. We diarise it and we tell you ninety days out. On synthetic voice: not for the brand-fronting read. We will use it for variable inserts inside a dynamic build where the stem is a place name, a price or a distance, and we will state in writing where it was used. A cloned voice delivering your brand promise is a legal exposure and, to anyone who listens to audio for a living, an audible one. Delivery specification: Recorded at 48 kHz and 24-bit, mixed with a true peak ceiling that stops platform limiters mangling the read, loudness normalised to the streaming target rather than a terrestrial one. Every mix gets a mono fold-down check, because smart speakers sum to mono and a wide stereo bed that sounded expensive on studio monitors collapses into mud on a kitchen counter. You receive the WAV master plus platform-encoded files and a 15-second cutdown that has been re-performed, not truncated. UK scripts go to Radiocentre for clearance before the session, not after it. Clearance changes wording, and recording first is how a studio day gets paid for twice. What we will not write: Spelled-out URLs. Promo codes longer than a syllable. Phone numbers. Two-hander comedy for a brand without the frequency budget to land the joke a second time. Jingles. Any call to action that assumes the listener has a free hand, which in car and kitchen listening is most of them. Not for: Podcast host-read briefs. Different craft, different people, and we will refer you rather than pretend. Brands wanting a voice reel on file with no campaign attached to it. Anyone whose script has to be approved by a committee that has not yet agreed on the single claim the spot is making. We will ask you to settle that question first, because settling it during a session costs studio hours and a performance that gets worse as the room loses confidence in what it is recording.