Fish Audio: Is It Free, What It Costs and What You Get

Fish Audio is an AI voice platform: you type a script, pick a voice and it reads it out, with emotion tags like [whispering] or [sighing] typed right into the text. It also clones voices, transcribes audio and has an API for developers. Is Fish Audio free? Yes, with an account: the Free plan costs $0 and gives you 8,000 credits a month, about 7 minutes of speech, for personal, non-commercial use only. Paid plans start at $15 a month with Plus, which adds enhanced voice cloning and commercial use; Pro is $100 and Max $999.

What Fish Audio is

Fish Audio is a web app at fish.audio, plus an API. The main tool is text-to-speech, and around it sit voice cloning, speech-to-text, Story Studio for multi-speaker stories, sound effects, audio separation and, more recently, image and video tools with lip-sync. The company behind it is Hanabi AI, and its models, published as Fish Speech under its own community license, are on GitHub.

Its newest model is Drama 3, which Fish Audio pitches as its most expressive one so far. You'll see the same idea all over the site: you don't just pick a voice, you direct it. The tag menu on the home page includes emotions like [angry], [sad], [excited] or [whispering], and special tags like [laughing], [sighing], [clear throat] or [long pause].

What you can do without an account

You can look around and type without an account. The home page has a demo with two voices, Sarah and Adrian, a box for your own text and a "Generate & play" button, and the home page says it supports more than 30 languages. I typed a short, boring script about a library's opening hours there, with a [pause] tag in the middle, to see how the tags sit in the text.

The web app, at fish.audio/app, opens without an account too. You can browse the menu, choose a model (Drama 3 was selected by default) and pick a voice. Under Default Voices there's a short list of voices with a "Fish Official" badge: Hannah, Adrian, Sarah, Ethan, Laura, Jordan and Selene. I picked Sarah, typed my script and pressed Generate Speech. That's where it stopped: a window titled "Unlock the Full Potential" asked me to sign in or sign up with email, Google or Apple. In the app, nothing plays until you sign up.

The other tab, Explore, is the community library: voices uploaded by users, which Fish Audio puts at more than 2 million. It's huge and very mixed, so if you're making something you'll publish, stick to a voice you own, or check a voice's license first (the voice library has a "Licensed" filter). The FAQ ties commercial use to "verified voices (that you own)."

Fish Audio pricing

These are the prices on fish.audio/plan as seen from the US, in US dollars. Outside the US, the page may show your local currency instead.

Fish Audio's FAQ puts a minute of speech at roughly 600 to 625 credits, but the plan cards promise fewer minutes than that math gives, so go by the minutes on each card. Credits reset every month, and unused minutes don't roll over.

When I checked, there was a launch offer for Drama 3: 50% off the first month on monthly plans, or 50% off the first year on annual ones, until November 12. That put Plus at $66 for the first year instead of $132. Offers like that come and go, so check the page before you pay.

The API, for developers

Fair warning: this part gets a bit technical. The API is billed separately from the monthly plans: it's pay as you go, with no subscription or minimum, and text-to-speech is billed by the size of your text, per million bytes. The paid S2 models cost $15 per million bytes and Drama 3 $50, and the docs say a million bytes is roughly 180,000 English words, or about 12 hours of speech. Until November 12, Drama 3 is billed at a launch price of $10. Speech-to-text costs $0.36 per hour of audio, and Voice Design, which creates a voice from a written description, is $0.01 per request.

One date to note if you already use it: the older S1 model is being retired on December 31, 2026, and requests that ask for it will then be served by an S2 model at the same price.

Voice cloning, Story Studio and the rest

Voice cloning gets top billing on Fish Audio's own site. The home page's title promises "free voice cloning," but the pricing page leaves enhanced voice cloning out of Free and puts it in Plus, and the professional voice slots grow with each plan. Fish Audio's home page promises cloning "in 15 seconds." Clone your own voice, or one you have permission to use.

Story Studio is for longer pieces with more than one speaker, like an audiobook chapter or a two-host podcast. Fish Audio's official tutorial on this page shows the whole workflow, from tags to cloning to a multi-speaker conversation.

The web app also has an Audio Separation tool for pulling voices and music apart. If you'd rather do that on your own computer for free, Ultimate Vocal Remover is the desktop app I'd use.

Who Fish Audio is for, and who should skip it

Fish Audio makes sense if you write scripts and want to direct how they're read: YouTube narration, audiobooks, game characters or a podcast with two voices. Developers get a simple pay-as-you-go API, and the Fish Speech models on GitHub are an extra if you like to tinker.

Skip it if you want to try your own script in the app without signing up, or if you need commercial audio for free. Neither is on offer here.

Fish Audio or something else?

Murf

Murf review →

If you want another text-to-speech studio to compare voices with, Murf is an online platform with a wide range of voices and languages. If what you really want is to change your own voice live, while gaming or streaming, Voice AI is built for that. And if your problem is a song rather than a script, Ultimate Vocal Remover splits vocals and instruments on your own machine, free.