Audiobook proofing. Turn listening notes into clear pickups.

Use a transcript to support your listening pass

An audiobook proofing pass checks a recording before it becomes the accepted production. A pickup is a passage the narrator records again to correct or replace audio. Polyglot can transcribe an authorized chapter locally on an Apple Silicon Mac, let you review text alongside playback and export a readable transcript with optional timestamps. That gives your listening notes a text reference; it does not approve the narration.

This workflow is useful for independent narrators, authors and producers who need to hand specific corrections back to an audio editor. Keep listening with the approved manuscript beside you. Polyglot does not automatically compare a book manuscript with its recording, detect every omission or create a pickup list. If your existing audio production tools already provide adequate transcript and review notes, a separate transcription step may not be necessary.

ACX’s Director’s Notes guidance recommends agreeing pronunciation and story details with the narrator and allowing time to listen carefully before submission. Those decisions remain human production work. The workflow below is an editorial suggestion, not an ACX requirement or endorsement.

Freeze the chapter and manuscript identities

Use a recording you own or are authorized to process. Confirm that the manuscript edition, narrator recording and any confidential pre-release material can be used in this workflow. Working access does not automatically permit publication of the transcript or a new translation. Keep review files with the intended production team and follow its storage and sharing rules.

Before importing, name the chapter audio revision, manuscript revision and pronunciation reference in a separate review document. Note whether the file contains one chapter or a larger assembly. A time such as 04:12 is only useful when the recipient knows which file it belongs to; it is not a location in the whole book unless you explicitly used that clock.

Work from the chapter cut agreed for proofing. Confirm that the recording uses a supported spoken language, then import it into Polyglot and review the generated text while listening. Audio-only input is supported. Keep your original recording and manuscript unchanged, and preserve earlier review exports when revised audio arrives.

Decide whether the text or the audio needs correction

For each suspected discrepancy, compare three things: the audible words, the approved manuscript and the generated transcript. If the narrator says the correct words but recognition changes them, correct the transcript. If the recording itself contains the wrong words, leave a clear audio correction request in your separate pickup list. Do not make the transcript match the manuscript and then treat that as proof that the recording is correct.

An approved adaptation is a third case. A rights holder may have agreed different wording for an audio edition. Refer that decision to the responsible producer instead of requesting a retake merely because the print text differs. Record the accepted wording and who approved it outside Polyglot.

This fictional example needs an audible-word check. A plausible transcript can hide the narration error, so searching the text alone is insufficient. The passage and review note are invented illustrations, not a processed customer audiobook.

Fictional chapter review
Audio: ch03-v02.wav
Manuscript: revision 4
Location: 04:12 in chapter
Anchor: “After the crossing”

Approved: “fifteen miles”
Heard: “fifty miles”
Draft: “fifteen miles”

Decision: audio pickup needed
Record the full sentence.
Do not close after text edit.

Give the narrator a usable pickup note

Export a reviewed transcript with timestamps if they help navigation. Treat each timestamp as a starting point, then verify the precise passage by listening. In a separate document, give each pickup a stable ID, chapter filename and revision, approximate time, a short spoken anchor, the issue and the approved replacement wording. Include enough surrounding sentence context for the narrator and editor to locate the right occurrence.

Keep pronunciation requests separate from spelling corrections. The transcript may spell a name correctly even when its pronunciation needs attention. Refer to the agreed pronunciation guide or an approved spoken reference, and ask the producer to resolve uncertain cases. Do not infer a pronunciation verdict from automatic text.

Agree the replacement span with your audio editor. A complete sentence may provide a more useful edit than a single isolated word, but the right span depends on the performance and recording. Track states such as requested, recorded, inserted and checked in your own document. These are production notes you maintain manually, not tasks or approvals managed by Polyglot.

Close pickups against the revised recording

After the narrator and editor make the changes in their recording and audio-editing tools, obtain a newly named chapter revision. Old timestamps may move when passages are inserted, removed or shortened. Use the pickup ID and spoken anchor to relocate each fix; do not carry an old timing reference forward as if it still identifies the same instant.

Import and review the revised chapter as a separate version if you need an updated transcript. Keep the earlier project and notes. Listen before, through and after each replacement to check the words, missing or repeated phrases and the transition into the surrounding narration. Mark a pickup checked only after hearing it in the revised file.

Finish with the agreed full-chapter listening pass, including its first and last words and the order of passages. A list with every requested pickup closed does not establish that no other error remains. Reconcile the final chapter filenames and review status before handing the audio onward.

Keep transcript review separate from audio mastering

A readable transcript cannot reveal every click, clipped breath, abrupt edit, performance change or level mismatch. ACX distinguishes edited audio from the final production master, and its Producer Toolkit points to separate audio submission requirements and analysis tools. Check the current requirements of your actual distributor with the appropriate audio tools and human review.

Polyglot does not record pickups, splice audio, master an audiobook or certify an ACX submission. Its output here is reviewed text that supports a manual production process. An SRT or WebVTT file is optional when you also need timed text; it is not a substitute for the finished audio deliverable.

Local media and transcript processing can suit pre-release review, but it does not replace permission or secure sharing decisions. Model downloads, user-controlled updates, optional content-free analytics and Pro licensing have separate network boundaries. Review the privacy information and your production agreement before using the workflow.

Keep exploring.