Turn Anything Into An Audiobook With Your Own Voice

My mom was never going to read 50 pages of my guides.

She has been asking me for months what I actually do all day. She wants to understand the AI stuff I post. But she is not going to sit down and read a stack of written guides, and she is not going to scroll through 300 videos. That is not how she takes in information.

So before her road trip last weekend, I cloned my own voice and turned every one of my AI guides into a single audiobook. Narrated by me. She listened to the whole thing in the car.

She texted me after and said she finally understood what I do. Then she asked how long it took. She assumed I sat in a room for five hours recording it. It took three minutes.

Then I stole my own idea. Now I turn long business research, AI newsletters, and every article I save and never open into audio I listen to while I walk or drive. This guide is the exact setup, the prompts I use, the real pricing, and the honest limits.

⚠ Read This Part First

The piece that makes this free is Fish Audio's S2.1 Pro Free API, and free access runs through the end of July 2026. That is days away as of this writing. After that window closes, the free developer model goes away and you move to the paid S2.1 Pro API, which is billed by usage. Set this up now, even if you only test it once. Getting the account and the voice clone done takes about ten minutes, and the voice you clone stays in your account after the free window ends.

Fish Audio makes text to speech models. Text goes in, a real sounding human voice comes out. Their newest model is S2.1 Pro, and they opened it up to developers for free while they collect feedback on it.

Free here means actually free. Here is what you get:

No credit card. You make an account, you generate a key, you use it.
No hard character cap. There is a Fair Use Policy instead of a fixed limit, so a full book length document is fine.
The same API as the paid version. Same endpoint, same quality. You change one value in the request to switch between free and paid later.
83 languages. So you can narrate in a language you do not speak.
Voice cloning included. This is the part people assume costs money. It does not right now.

Two things to know before you build anything serious on it. If your product makes more than $1M in annual recurring revenue, Fish Audio asks you to contact them before using S2.1 Pro Free. And there is no uptime or latency guarantee on the free tier. It is best effort, not a contract. Fine for your audiobook. Not something to put a paying customer on.

Plan	Price	What it is for
S2.1 Pro Free API	$0 through end of July 2026	The developer offering this guide uses. Voice cloning, 83 languages, no card.
Free consumer plan	$0	The website only. This plan does not include API access, so it is not the same thing as above.
Plus	$11/mo	Consumer plan for people using the site rather than the API.
Pro	$75/mo	The heavier consumer plan.

That second row trips people up, so I want to be direct about it. The free consumer plan and the free API are two different things. Signing up for the $0 consumer plan does not get you API access. The free API is the separate S2.1 Pro Free developer offering, and that is the one with the July deadline on it.

You do not need to know how to code for this. You need an account, a key, and a chat with Claude. If you have Claude Code, it will do the whole last step for you.

Step One

Get your account and your API key

Go to fish.audio and make an account. Then open the API keys page in your dashboard and create a new key. Copy it right then, because most platforms only show you a key once.

Save it somewhere private. That key is tied to your account, so treat it like a password. Do not paste it into a public chat, a public repo, or a screenshot.

Step Two

Pick your voice, or clone your own

You have two options here, and I use both for different things.

Clone your voice. Fish Audio can build a voice from a short sample. Record yourself reading a paragraph or two in a quiet room, upload it, and you get a voice model back. Read normally. Do not perform it. If you read the sample in a big announcer voice, everything it makes later sounds like that too. Each voice you create gets an ID, and that ID is what you hand to the API later.

Or use a library voice. Fish Audio has a community voice library with a lot of options, including documentary style narrator voices. This is what I use for my own listening. I cloned my voice for my mom because the point was that it was me. For turning a newsletter into a podcast for myself, I would rather have a narrator. Listening to hours of my own voice would be strange.

Step Three

Have Claude rewrite the document for the ear

This is the step people skip, and it is the one that decides whether the result sounds like an audiobook or a robot reading a spreadsheet.

Written documents are full of things that make no sense out loud. Headings. Bullet symbols. Links. Numbers in tables. If you dump a raw document straight into a voice model, you get "bullet, bullet, h t t p s colon slash slash" read out at you. So you have Claude convert it into narration text first.

Upload your document to Claude and paste this in.

Prompt 1, Document To Narration Script

I uploaded a document. Rewrite it as a script that will be read out loud by a text to speech model. This is going to become an audiobook, so it has to work for the ear, not the eye. Rules, follow all of them: 1. Output plain narration text only. No headings, no titles, no section numbers, no markdown, no asterisks, no bullet symbols, no dashes used as list markers, no tables. 2. Where the original used a heading, replace it with a short spoken transition instead. Something like "Next, the part about pricing." Never read a heading as a label. 3. Where the original used a list, turn it into a spoken sentence. "There are three things that matter here. First... Second... Third..." Never read symbols aloud. 4. Remove every URL, link, footnote marker, and citation number. If a link mattered, say the name of the source out loud instead, like "according to their pricing page." Never spell out a web address. 5. Spell out anything that would be read wrong. Write numbers, dates, currency, and units the way a person would say them. Write "twenty twenty six" not "2026". Write "eleven dollars a month" not "$11/mo". Expand abbreviations the first time, so "annual recurring revenue" instead of "ARR". 6. Break long sentences into short ones. If a sentence would make a person run out of breath, split it. 7. Keep all the real information. Do not summarize, do not shorten the substance, do not add anything that was not in the document. Only change the form. 8. Use paragraph breaks where a narrator would naturally pause. 9. Do not add an intro, an outro, or a sign off unless I ask for one. Start with one short spoken opening line that tells me what I am about to listen to. Then give me the full narration text and nothing else. No preamble, no notes, no explanation of what you did.

Read the first few lines of what comes back before you send it anywhere. If it still has a heading or a stray symbol in it, say "line three still has a bullet, fix it" and it will.

Then send it to Fish Audio

If you use Claude Code, or any AI tool that can run code and make web requests, this is one more prompt. Paste your narration text and your voice ID in and let it handle the request.

Prompt 2, Send It To Fish Audio

I want to turn the narration text above into an audio file using the Fish Audio API. Write and run a small script that does this for me. Details: Endpoint: POST to https://api.fish.audio/v1/tts Auth: Bearer token in the Authorization header. My key is in the environment variable FISH_API_KEY. Read it from there, never print it, and never write it into the script itself. Model: send the header model with the value s2.1-pro-free Body: JSON with the narration text, reference_id set to [MY VOICE ID], and format set to mp3. Requirements: 1. Split my text into chunks at paragraph breaks before sending, so no single request is enormous. Never split in the middle of a sentence. 2. Send the chunks in order, then join the returned audio into one mp3 file in that same order. 3. Save the finished file as [FILE NAME].mp3 in this folder. 4. If a request fails, wait a few seconds and retry it up to three times before giving up. This is a free tier with no uptime guarantee, so occasional failures are expected. 5. Tell me the final file name and how long the audio is when you are done. Before you run it, show me the plan in three lines so I can confirm.

Two things to fill in. [MY VOICE ID] is the ID of the voice you cloned or picked in step two. [FILE NAME] is whatever you want the file called. Then drop the finished mp3 into your phone, or into any podcast app that lets you add your own files, and it plays like anything else in your library.

If You Do Not Use Claude Code

You can still do steps one and two, and use Fish Audio's own site to generate the audio from the narration text Claude wrote. You lose the automatic chunking and joining, so you will paste in sections and stitch them yourself. The narration prompt above is still the part that makes it sound good, and that works no matter how you generate the audio.

The trick is not to think of this as making audiobooks. Think of it as everything you told yourself you would read later. Here is what I actually run through it.

Long business research. Market reports, competitor breakdowns, anything my AI produced that is 20 pages long. I would never sit and read those. I will absolutely listen to one on a walk.
AI newsletters. I subscribe to more than I can keep up with. Once a week I take the ones I did not open, have Claude combine them into one narration script, and turn it into a single episode.
Saved articles. Every person has a read later folder that is really a never folder. This empties it.
Your own guides and documents. This is the one for other people. If you have written anything you want someone to actually take in, an onboarding doc, a family history, a set of instructions, audio gets it consumed when reading would not.
Course notes and study material. Turn a chapter into audio and review it on the drive to class or work. Hearing something after you have read it is a real second pass.
Meeting docs and long email threads. Prep for a meeting on the way to the meeting.

One habit that made this stick for me. I do not do it one file at a time. Once a week I gather everything I did not get to, have Claude turn the whole pile into one narration script, and make a single long file. One thing to press play on beats seven separate ones.

I am not going to sell you this without the other side of it.

The free window ends with July 2026. This is the big one. S2.1 Pro Free is free through the end of July 2026, and that is close. After that, the paid S2.1 Pro API is still there and still good, but you will be paying for usage. Set it up now while the free version is live, and decide later whether it is worth paying for. You will know by then, because you will know how often you actually used it.
No uptime or latency guarantee. The free tier is best effort, not contractual. Requests can be slow or fail. That is why the second prompt above builds in retries. It is completely fine for your own listening. Do not put a paying customer's experience on it.
Over $1M in annual recurring revenue means you need to ask first. Fish Audio asks products above that line to contact them before using S2.1 Pro Free. If that is you, send the email before you build.
Cloning your own voice is great for personal use. Be thoughtful about the rest. Only clone a voice that belongs to you, or one you have clear permission to use. And think about where a cloned voice goes. A voice narrating guides for my mom is one thing. A voice saying something you never said, sent to someone who thinks it is really you, is a different thing entirely. My rule is simple. Clone my own voice, tell people when it is a clone, and never use it anywhere that a person might act on it thinking it was live.
It is a very good machine voice, not a performance. It will not do the emotional beats a real narrator does. For information, it is more than good enough. For something that lives or dies on delivery, record it yourself.

The voice clone is the part people react to. It is not the part that mattered.

What mattered is that my mom got the information in the format she was actually going to take it in. She did not need a shorter guide. She needed a different one. Written was never going to work for her, and no amount of me making it prettier was going to change that.

Then the same thing happened to me. I have been telling myself for years that I would read the research, the newsletters, the saved articles. I did not. Not because I did not care, but because reading requires me to sit still, and I do not sit still much.

The Real Win

I get through roughly twice as much information now, and I did not add a single hour to my week. Reading time turned into walking time and driving time. Those hours were already there. They were just being spent on nothing.

That is the whole idea. Not fancier content. The same content, in a format that fits the hours you already have.

Start with one thing. Pick the longest document sitting in your saved folder right now, run it through the two prompts above, and listen to it tomorrow. If you like it, you have a system. If you do not, you spent ten minutes. And do it before August, because the free version has a date on it.

Turn AnythingInto An AudiobookWith Your Own Voice

Turn Anything
Into An Audiobook
With Your Own Voice