AI narration has a quirk that human narration doesn't: it is too clean. Between lines, there is typically nothing at all - not a quiet room, not a breath of air, just perfect digital silence. To a listener that sounds fine. To an audiobook platform's automated checks, it can look like a red flag.
This week we added a set of tools to Radio Novel that take care of that for you. Here is what's new, how it works, and - importantly - what it does not do.
Why "perfectly silent" is a problem
Audiobook platforms publish technical requirements for the files you submit, and they check them automatically before a person ever listens. The common ones are:
- Loudness - the overall level of the recording has to land inside a set window.
- Peaks - no moment can be louder than a set ceiling.
- Noise floor - the quietest parts of the recording have to be quiet, but not empty. A real recording always has a little room sound underneath. A file with true digital silence between lines doesn't, and that can get it rejected.
- Room tone at the start and end - a short stretch of that quiet room sound before the first word and after the last.
An author using Radio Novel shared how long it took them to work out why their AI narration kept failing these checks, and how much editing it took to fix it by hand. That feedback is what started this work: the right answer is for Radio Novel to handle it for you.
What's new: Master Audio Settings
Open any completed audio production and you'll find a new Master Audio Settings button. It controls how your final mix is finished:
- Room tone - a very quiet, natural-sounding background fills the gaps and the start and end of your recording, so there is never dead silence. You choose how much goes before and after the narration.
- Loudness target and peak ceiling - set the overall level and the loudest allowed moment. The defaults are set to the published specs of ACX, one of the major audiobook platforms, as we understand them. Requirements differ between platforms and change over time, so check the ones you are submitting to.
- Compression - optional, off by default, to even out louder and quieter passages.
- Pause between lines and between paragraphs - control the pacing of your finished audiobook, with the defaults matching what Radio Novel has always used.
Press Save & Recompile and your final mix is rebuilt as a new version. Your earlier version stays available to download. The rebuild reuses your saved tracks, so it normally doesn't use up any regenerations.
New productions get these defaults automatically. Existing productions keep their current audio until you recompile them.
See exactly how your mix measured
The same window now includes a Last mix, as measured report: your loudness, peak level and noise floor, whether the recording contains any digital silence, and how much room tone sits at each end - each with a clear pass or fail against the published targets.
Below that, a Pacing summary shows how your recording breaks down between the first and last word: how much is dialog, how much is sound (music, ambience and effects), and how much is silence, to a tenth of a percent. It also counts your pauses by length and points to the longest one, so you can see whether any pause might drag.
About the room tone
The room tone isn't a looped recording, which could repeat in a way you can hear or a computer can detect. Radio Novel generates it fresh for each production, modelled on the sound of a real, quiet vocal booth. It sits far below the narration - quiet enough that at normal listening volume you won't hear it at all. Its job is to be present, not noticeable.
We also give it its own track. When you download all of your tracks, room_tone.mp3 is included alongside your narration, sound effects, ambience and music, all perfectly lined up with the final mix. If you edit in a program like Audacity, you can see it, mute it or swap it for your own.
Your individual tracks are never altered. The finishing only touches the final mix.
Important: this is not a guarantee
Please read this part carefully.
Meeting a publisher's sound standards does not relieve you of the obligation to meet that publisher's publishing policies on AI disclosure, and it does not guarantee acceptance.
Radio Novel's checks are a guide to whether your audio meets common technical targets. They are not an approval from any platform. Each audiobook publisher and distributor sets its own rules, and they change. Many have specific policies about AI-generated or AI-narrated content, including how it must be disclosed. Whether your audiobook is accepted is decided by the platform, based on its current requirements and its review, and a technically perfect file can still be declined.
Before you submit, read the platform's current technical requirements and its current policy on AI narration, and follow them, including any disclosure it asks you to make. If you are unsure, ask the platform directly. You are responsible for what you submit and for complying with the terms you agree to when you publish.
A few smaller improvements
- Quicker navigation. Every audio and video production in your lists now has direct links to its Story Workspace, its Audio Production and its Video Production, whenever those still exist. Your Story Workspace list has the link too.
- Tidier lists. We fixed uneven spacing between productions in your Audio Productions list.
- Menus stay on top. The "..." menus on your Projects and Audio Productions pages used to slip behind the items beneath them. They now always show in front.
Try it
Open a completed audio production, choose Master Audio Settings, review the defaults, and press Save & Recompile. When it finishes, open the window again to read your report.
As always, we'd love to hear how it goes. If a platform's checks flag something these tools should have caught, tell us - that is exactly how this feature started.
Sign in to leave a comment.
No comments yet - be the first to share your thoughts.