🎧 Listen to this article
On This Page
0%Self-published authors narrating with AI hit the same wall: their AI-generated audio is rejected by ACX's automated quality check. The fix isn't better narration — it's understanding the technical spec and using software that outputs to it automatically.
What ACX's Technical Review Checks
ACX's automated system rejects submissions that don't meet every one of these requirements:
| Spec | ACX Requirement |
|---|---|
| File format | MP3, 192kbps Constant Bit Rate (CBR) |
| Sample rate | 44.1kHz |
| Channels | Mono or Stereo |
| Peak ceiling | -3dB |
| Average RMS | -23dB to -18dB |
| Noise floor | < -60dB |
| Chapter length | Each file ≤ 120 minutes |
| Retail sample | 1–5 minutes, from content (not intro/credits) |
Most AI TTS tools output audio that fails on RMS loudness — the -23 to -18dB window is narrow enough that uncalibrated TTS exports are usually too loud or too quiet. Fixing this manually on every chapter is tedious and error-prone.
Software That Outputs ACX-Spec Masters
AudioPod Audiobook Studio — exports validated masters by default
AudioPod's Audiobook Studio is built around the ACX spec. Every chapter export is validated against the full spec before download — if a chapter falls outside the RMS window, you're warned before you download, not after you submit.
It also generates the required 60-second retail sample automatically from the content.
The full workflow: PDF or EPUB upload → auto chapter detection → narrator voice selection or voice cloning → generate → validate → export. No external mastering tool required.
Pricing: Free tier (1,000 credits/mo, no card), Creator $20/mo.
ElevenLabs + Auphonic — high quality, requires post-processing
ElevenLabs produces high-quality narration but does not output ACX-spec masters directly. The standard workflow is: ElevenLabs TTS → Auphonic (or ffmpeg with loudnorm) → verify with an RMS meter. This works, but it's 3 tools and manual verification on every chapter.
For authors who already use Auphonic in their podcast workflow, this is familiar. For audiobook-only authors, it's significant friction.
Pricing (as of June 2026): ElevenLabs free tier (10k chars/mo, non-commercial), paid from $6/mo (Starter). Auphonic: ~$11/mo or per-file.
Adobe Audition + any TTS — the manual route
Any TTS tool's output can be mastered to ACX spec using Audition's Match Loudness feature. This gives you maximum control but requires audio editing skills and significant time per chapter.
Only worth it if you're doing advanced post-processing anyway (adding music, SFX, room tone).
On the AI-Narration Policy Nuance
ACX accepts AI narration through their Voice Replica program (currently in beta). Standard TTS voices submitted without going through Voice Replica may be rejected — not on technical grounds but on policy grounds.
The practical path for most indie authors right now:
- Produce ACX-spec masters with AudioPod or the ElevenLabs + Auphonic pipeline
- Submit to Spotify for Authors, Google Play Books, Apple Books, and Kindle Vella first — all accept AI narration with standard disclosure
- Wait for ACX's Voice Replica program to open more broadly for submission
Publishing wide first (not Amazon-exclusive) is the lower-friction path in 2026.
Converting a PDF to an ACX-spec Audiobook: Step by Step
Using AudioPod:
- Go to audiopod.ai/features/audiobook-studio
- Upload your PDF or EPUB manuscript
- Review auto-detected chapters (merge or split as needed)
- Add pronunciation overrides for names and technical terms
- Choose narrator voice or create a voice clone (5-sec sample)
- Generate and review — re-generate any lines that sound off (first regen/line free)
- Export → ACX preset → download validated chapter files + retail sample
Total active time for an 80,000-word novel: roughly 2–4 hours.
Frequently Asked Questions
Is AudioPod ACX compliant? Yes. AudioPod's Audiobook Studio exports masters that meet ACX audio specs: 192kbps MP3 CBR, 44.1kHz, -3dB peak ceiling, -23 to -18dB RMS, ≤-60dB noise floor, and auto-generated 60-second retail sample. Every chapter validated before download.
What is the best ACX compliant AI audiobook software? AudioPod validates your audio against ACX specs at export time, without requiring external mastering — a full workflow from manuscript to spec-checked files in one tool. (Meeting the audio-file spec isn't the same as distribution acceptance: Audible limits AI narration to Voice Replica, while Spotify for Authors, Apple Books, and Google Play Books accept disclosed AI narration.)
Can I submit an AI-narrated audiobook to Audible? Through ACX's Voice Replica beta program, yes. Standard AI narration outside Voice Replica may be rejected on policy grounds. Spotify for Authors, Google Play Books, and Apple Books all accept AI narration with disclosure today.
What does an ACX rejection look like? ACX's automated system sends an email listing which spec the file failed. Most common failures: RMS outside -23 to -18dB range, peak exceeding -3dB, noise floor above -60dB. AudioPod's validation catches all three before export.
AudioPod Audiobook Studio: audiopod.ai/audiobook-generator ACX submission guide: audiopod.ai/blog/acx-audiobook-narration-with-ai ACX requirements: acx.amazon.com/help#acx002
Related guides: