Fanfic is probably the only literature that exists, in full, only for the eyes. Professional narrators don't read fanfic — there's no publisher to pay one. Meanwhile readers are out there grinding through 300- and 500-thousand-word epics on AO3, and by the end of a long sequel everyone's eyes hurt. I've heard "I wish this was audio" a lot — and ten minutes of work in a service like ours now closes that gap.
Here's how it actually goes, on a typical 70–100k-word fic.
What to gather before starting
The text itself, as a file. AO3 has a Download button in the top right of every work that gives you EPUB — that's the ideal source, full chapter structure included. FFN's export options are limited; easiest there is to copy text into a notepad and save as .txt. If the fic is split across many chapters on the site, stitch them into one file before uploading.
An account on an AI narration service. Could be ours, could be a competitor — for short fics the differences are small.
Headphones. Even cheap earbuds. Speakers smooth artifacts, and your impression of chapter one in headphones is the more honest one.
The actual steps
Upload the file. Then go through this in order.
Check how the service split the text into chapters. If it gave you one monolithic block, your source doesn't have chapter markers. Two minutes in a notepad fixes it: add # Chapter 1, # Chapter 2 lines before each chapter break. Without structure, the player can't jump and listening gets harder.
Review the character list the service detected. Auto-detection works by scanning names and dialogue tags. Fandom fics with rare names sometimes confuse it. Walk through the list and make sure every canon character has something reasonable assigned.
Cast voices for the main pairing or core characters. This is the important part, and on a fic about a specific couple I'd spend ten minutes here, not five. The voice you hear Snape in, or Loki, or Aziraphale, is half the success. Side characters are fine on automation.
Start the render. A 70k-word fic finishes in ten to twenty minutes. Time enough to make tea.
Voices for canon characters
There's no universal table, but a few patterns hold across fandoms.
Brooding adult men — Snape, Loki, post-timeskip Viktor, older Allen Walker — low male voice, measured pace, a touch of cold in the style hint. "Cold, detached" is usually enough.
Teen protagonists — slightly higher than mid-range, male or female by canon, moderate pace, more emotion than baseline.
Wise mentor figures — Dumbledore, Gandalf, Iruka, Ironwood when written kindly — older male, slow, warm. The Gandalf template still works, even outside its own fandom.
Villains. Don't reach for an "evil voice." Seriously. Tom Riddle, Aizen, end-arc Ozpin all read best in an even, calm, slightly cold voice — no theatrics, no laughter. Cold restraint is more frightening than shouted rage.
Most decent services give you five to ten male voices and the same number of female. I'd spend three minutes listening to each on a short phrase before committing — it's worth the time.
Where this usually breaks
No chapter breaks in the source. Already covered above — two minutes in notepad to add headers.
Dialogue without attribution. If a chunk of replies skip "she said," "he answered" tags, AI starts mixing speakers up. In short scenes that's fine; in long two-character dialogues it's audibly off. For important scenes you can hand-add attributions; figure five minutes per chapter.
Foreign names. Especially from East Asian fandoms — manga, anime, K-pop. Transliterations of names like Satoru, Kyojuro, Taehyung, or Yuri Katsuki can land badly in English models. Hack: listen to the first two chapters, and if a name comes out warped, write a phonetic spelling in the source (or use stress markers if the service supports them).
Long author's notes. A lot of authors front-load chapters with "A/N: thanks to my beta," disclaimers, or "sorry, this got angsty instead of fluffy." AI reads these as part of the chapter, and in audio it's jarring. Strip them out before upload, or move them to a separate section.
Rights and the unwritten rules
Fanfic is copyright's grey zone, and there are unwritten fandom norms that you'd do well not to break.
For personal listening — no problem at all. You narrated your favorite fic for yourself, you're listening on the train, nobody's coming after you. Functionally identical to printing the text for your own reading.
For public distribution — different story. If you want to put the audio version on YouTube or in a Discord, three things to keep in mind. First, contact the author beforehand. Most fanfic authors are happy about fan adaptations as long as they're credited; some ask for explicit approval. Second, credit the author in every chapter, with a link to the original. Third, don't monetize. Donation-based ("if you enjoy, support") is generally accepted; outright sales of fan-made audio are not.
This isn't law, it's the fandom code. Breaking it gets your channels reported and sometimes pulled.
Long fics are their own thing
Half-million-word epics aren't "narrate in ten minutes," they're a real project. Render takes hours, sometimes overnight.
Decent services (we're one) prioritize the early chapters — meaning your first hour of audio is ready in thirty minutes, while the remaining twenty hours cook in the background. Don't wait for the full thing — start listening, the tail catches up.
Casting for long fics matters more, especially for the main pairing and recurring side characters. For minor characters that show up in three out of a hundred scenes, I wouldn't bother — let the automation sort them.
Short version
Ten minutes for a one-shot, a couple of hours of attention for an epic. Your first audio fanfic is the hardest one — after that they stack up fast. I have dozens in the commute playlist now, and I've stopped noticing it isn't a person reading.