Public beta · First accepted audio file free Private upload · Private delivery
← All guidesAUDIOBOOK PRODUCTION / TECHNICAL CHECKLIST

Preparing AI-Generated Voiceovers for Audiobook & Broadcast Technical Specifications

Long-form narration needs more than a pleasing voice. File encoding, chapter layout, level consistency and noise floor matter when preparing work for a specific distributor.

Start with the destination specification

Do not infer delivery requirements from the source file. A 24 kHz or 48 kHz recording is not inherently defective or evidence of AI generation. ACX currently calls for a 44.1 kHz MP3 at 192 kbps or higher constant bit rate. A 48 kHz source therefore needs a deliberate conversion for that destination; a 24 kHz source cannot regain missing high-frequency detail by upsampling.

VoxErase currently accepts one MP3 or WAV up to 50 MB per upload and delivers one processed file for accepted requests. We do not certify ACX, Audible or broadcast compliance.

Measure the finished chapters

ACX states an RMS range of -23 to -18 dB, peaks below -3 dB and a noise floor below -60 dB RMS. It also asks for consistent channel format, tone, pacing and sound quality across chapters. Leave controlled room tone at chapter boundaries and inspect edits for clicks, plosives and abrupt noise changes.

These requirements can change. Check the current distributor specification and measure the final encoded export, not just the editing session. ACX Audio Lab can help assess several metrics but does not replace listening or all noise checks.

CBR, VBR and MP3 headers

A variable-bit-rate MP3 may contain a Xing frame, but a Xing or Info byte string alone is not a complete CBR/VBR diagnosis. Inspect the encode settings and verify the final file with a reliable media probe. If targeting ACX, export 192 kbps or higher CBR as specified.

For very long projects, plan the one-file beta test around a representative passage. Contact us about chapter batches; the current public beta does not include an active monthly plan.

Primary references

VoxErase delivers one processed audio file for accepted requests. Processing results vary. We do not guarantee removal of embedded watermarks, a detector score, or platform acceptance.

Explore the service, read how intake and delivery work, or submit your first beta file.