Producing an Audiobook
Chapterized long-form narration with a restrained music bed, delivered to pass a platform's audio check.
Tool profile
- Kind
- guide
- Vendor
- —
- Form
- 流程
- Platforms
- 通用
- Price
- 不适用
- Jobs
- 流程
- Verified
- 2026-10-31
Related
Narration is the only through-line
In an audiobook, music is a guest. A narration runs for ten or fifteen minutes without a break, and any bed with a melody, a lyric or a clear pulse competes with the stress of the sentence. Ask three questions first: is there character dialogue, does the chapter need a transition, and does the platform allow music at all. If the platform checks a mono voice track, keep the body dry and put music only at chapter edges.
Three places music can go
| Position | Length | Approach | Common problem |
|---|---|---|---|
| Chapter opening | 5–15 s | One statement of the theme, after the chapter title | Steals the first spoken line |
| Chapter ending | 10–30 s | Close without fading to nothing | Cuts the narrator's tail |
| Under the body | Usually none | If required, 25 dB or more below the voice and without melody | Long-form fatigue, failed QC |
Decide the position before writing. Reuse one theme at chapter openings so the book feels whole; a new cue for every chapter makes chapter three feel like a different book.
Recording, editing and seams
Record one chapter at a time and redo fluffs on the spot. Keep three to five seconds of room tone per chapter and use it to fill edits, or the seam will alternate between silence and hiss. Noise reduction should remove steady noise only; do not squeeze the floor into a pumping void. Keep breaths and sibilance. Distinguish characters by distance, delivery and slight timbre, not pitch shifting. Put music on its own track and drop it 2–3 dB between phrases, never inside a phrase. Reaper, Pro Tools or Audition all work; keep batch processing and loudness reports on file.
Delivery checklist
- One file per chapter, named with book code, chapter and version
- Mono 44.1 kHz, 24 bit WAV, plus the MP3 the platform asks for
- Half a second to one second of room tone at the head, one to five seconds at the tail, no long gaps in the body
- Chapter loudness on the platform target, peak below its ceiling
- Acceptable noise floor in quiet passages, no metallic residue after cleanup
- First and last lines complete, no swallowed words or half breaths
- Music and dry versions equal length and same start
- A pronunciation sheet for names, places and specialist terms
Where it breaks
Chapter levels drift and the listener adjusts volume every chapter. Over-reduction thins the sibilance until the voice sounds like it is behind paper. Music fades at the end of a phrase and the next line surfaces from underwater. Pitch-shifted characters sound like a different announcer. Chapters start at different points and the seams show in continuous playback. Only WAV is delivered, the platform wants MP3, and the transcode overshoots. A bed with lyrics fights the narration. The whole book arrives as one file, so the platform cannot chapterize it and the listener cannot jump.
Listen in the library
All tracks from midicn-lib — inline badge is the license tier (C1 commercial / C2 non-commercial / C3 study only).
Along the same line
Sources
- ACX 音频提交要求（Audible/ACX 官方）— 响度、峰值、底噪与文件格式的验收口径。
- AES TD1008《Recommendations for Loudness of Internet Audio Streaming and On-Demand Distribution》（2021）— 语音类内容的响度目标。
- Adobe Audition 官方文档 — Match Loudness 与 Diagnostics 面板的测量与批处理用法。