Making an Audiobook: From Recording to Finished Chapters
Three hours of recorded book later, playback reveals uneven chapter volumes, coughs and flubs at boundaries, and twenty files in random order. Audiobook production is technically simple — split, clean, unify loudness, merge — as long as the four stages run in order.
Two sources, one pipeline
Self-recorded narration: phone or microphone capture with room noise and volume drift to handle. Synthesized speech: TTS-generated chapters, clean but with uneven loudness across chapters and stiff pauses. Both feed the same post-production chain: chapter confirmation → trimming and cleanup → loudness unification → merge and export.
The source only changes the workload of stage one: recordings need noise reduction and flub removal; synthesized speech usually skips straight to loudness.
Split into chapters first
Organize as one file per chapter (or per subsection) with zero-padded names (01, 02 …). Never merge before fixing — a broken spot inside a merged three-hour file means redoing everything, while chapter files are independently reworkable.
Pause between chapters while recording and you get chapter files for free; if you recorded one long take, cut it into chapters with a trim tool before continuing.
Cleanup: removing flubs and dead air
At each chapter boundary: cut coughs, page turns and equipment bumps; keep about half a second of breathing room before the narration starts and after it ends — cutting tight to the words sounds abrupt. Inside chapters, trim dead pauses beyond three seconds and obvious retakes.
Do not line-edit every sentence — audiobook listening values continuity and evenness, and the occasional natural pause is rhythm. Only remove clearly broken material.
Loudness unification: the biggest variable in listening comfort
Across three hours, nothing hurts more than loudness: volume jumping between chapters, narration dropping during character dialogue, listeners constantly adjusting. Unify every chapter to one target loudness; in-chapter dynamics settle naturally around it.
Order matters: unify chapters first, then merge — merging first and averaging afterward is not the same as every segment being right.
Merge granularity for publishing
Platform-dependent: audiobook platforms with chapter lists take per-chapter files (upload in order, the platform sequences them); single-file scenarios (USB drive, cloud folder for elderly listeners) merge into one book or parts (every 5–10 chapters, 1–2 hours per file, long enough to feel complete, short enough to seek).
Merging again requires matching parameters: unify sample rate and format during chapter processing and the merge is pure stream copy, done in seconds.
Format and bitrate: subtraction for voice-only content
Audiobooks are pure speech — bitrate can drop hard: AAC 64 kbps or MP3 96 kbps is fully sufficient, indistinguishable from 192 kbps at three times the size. A three-hour book at 64 kbps AAC runs about 85 MB, friendly to cloud storage and phones.
Format choice: mp3 for the widest compatibility (old devices, car systems), m4a/AAC for smaller files on modern devices. The full selection logic lives in the audio format guide; for voice-only content, "sufficient" wins.