Disclosure: some links on this page may be affiliate links; we may earn a commission at no extra cost to you. Read the disclosure.

By the toolist.ai editors. Researched from Fish Audio's own pages and hands-on tests published by others, not hands-on tested by toolist.ai. Updated October 14, 2026.

Fish Audio review: pricing, features and what testers found

Fish Audio turns text into expressive speech you direct in plain words, clones voices from a short clip and hosts 2,000,000+ community voices, in a web app, iOS and Android apps and a pay-as-you-go API. The free plan gives up to 7 minutes a month for personal use; Plus costs $15 a month ($132 a year) for up to 200 minutes, and the API charges $15 per million UTF-8 bytes of text.

Our take

Fish Audio suits creators and developers who want expressive speech they can direct in plain words, cloned or community voices, and low prices: Plus costs $15 a month ($11 a month billed yearly) for up to 200 minutes of speech, and the API charges $15 per million UTF-8 bytes, which Fish Audio puts at about 12 hours of speech, with S2.1 Pro free to developers under fair use through November 30, 2026. On Artificial Analysis's blind-vote leaderboard on October 14, 2026, its new Drama 3 Preview ranked 10th of 97 models and S2.1 Pro 18th, behind ElevenLabs' two Eleven v4 models (1st and 2nd) and Cartesia's Sonic 3.6 (4th), though Drama 3 Preview had the top score on the same site's pronunciation test. Zapier's Miguel Rebelo picked it for its 2,000,000+ community voices but found their quality uneven and emotion tags hit or miss depending on the voice, and Feisworld's Fei Wu found its voice cloning quicker to set up than any other tool she had tried. The catches: the pricing page contradicts itself on how many minutes the credits buy, the help center calls all purchases non-refundable although the Pro card promises a 7-day money-back guarantee, the free plan is for personal use only, and the community library has carried voice clones uploaded without the speakers' consent.

Based on Fish Audio's published plans and the hands-on tests linked below, not on a test of our own.

Good for

  • Creators who want expressive, directed voiceovers for YouTube, audiobooks, audio drama and games at a low monthly price: Plus is $15 a month for up to 200 minutes of speech
  • Finding a character voice in a 2,000,000+ community library, or cloning your own voice from a short clip
  • Developers who want a text-to-speech API priced per byte of text, a free S2.1 Pro model for prototypes through November 30, 2026, or open model weights to run themselves

Watch out for

  • The pricing page contradicts itself on what the credits buy: its FAQ says a minute of generation costs roughly 600 to 625 credits, which would make Plus's 250,000 credits 400 minutes or more, but the plan cards say up to 200 minutes on Plus and 1,620 on Pro (about 1,250 credits a minute) and only 6,250 on Max, although Max's 25,000,000 credits are 12.5 times Pro's (4,000 credits a minute), and the page's own data gives the free plan 1 hour against 7 minutes on its card. The help center says speech credits are charged per byte of text and that costs vary by model and may change. This record counts the minutes on the plan cards, the lowest of the page's figures, so Max may buy more speech than it lists.
  • Refunds conflict: the Pro card promises a 7-day money-back guarantee, but the terms say fees already paid are not refunded except as the agreement expressly provides, and the help center calls all purchases non-refundable, accidental ones included, adding that a processing fee of about 15% may apply when support reviews a case.
  • The free plan is for personal, non-commercial use only, and anything published from it must credit 'Fish Audio' or 'fish.audio' in the title or description; audio made during a paid subscription stays usable commercially, but audio made before or after one never is.
  • Plan credits refill monthly on your billing date and do not roll over, unused credits lapse when a cancelled or downgraded plan ends, and plans renew automatically at the then-current non-promotional price with no prorated refund; API credits are sold separately from plan credits.
  • A Drama 3 launch offer runs until November 12, 2026: 50% off a new Plus, Pro or Max plan's first month (monthly) or first year (yearly), such as Plus at $66 for its first year, and Drama 3's API at $10 per million characters instead of its $50 list price. This record lists the regular prices, which apply at renewal. The API pricing page does not list Drama 3 yet and prices the other models per UTF-8 byte, not per character.
  • Anyone can add a voice clone from a 10-second clip: Zapier's tester found fans had uploaded TV and video game characters without the actors' consent, and TechTimes reported in July 2026 that UK voice actors and their union Equity had complained about unauthorised clones, and that Fish Audio's CEO had told TechCrunch removal is now automated once a creator proves ownership; the published model weights can still clone voices where no takedown reaches.
  • The terms (effective August 18, 2024) let Fish Audio use your content and usage data to train its AI models; the help center says API customers' data is not used for training by default, but requests to the free S2.1 Pro API may be used to improve the model, and zero data retention is an Enterprise feature.

What Fish Audio does

Price

Free plan: 8,000 credits a month (up to 7 minutes of speech), 500 characters per generation and 3 public voice slots, for personal, non-commercial use only and with credit to Fish Audio; developers can also call the S2.1 Pro model through the API free, under fair use, through November 30, 2026.

PlanBilled monthlyBilled yearly, per monthIncludesPrice per minute of generated speech
Free free free No card needed: 8,000 credits a month, which the card puts at up to 7 minutes of generation (the page's data says 1 hour); up to 500 characters per generation, 3 public voice slots and standard generation speed; no enhanced voice cloning and no commercial use, and published audio must credit Fish Audio –
Plus $15 $11 $132 billed yearly (the page calls it 3 months off). 250,000 credits a month, which the card puts at up to 200 minutes of generation; up to 15,000 characters per generation, unlimited public and 10 private voice slots, 1 professional voice slot, Voice Design, priority generation on the latest models, enhanced voice cloning and commercial use $0.075
Pro $100 $75 $900 billed yearly. 2,000,000 credits a month, which the card puts at up to 1,620 minutes (27 hours in the page's data); up to 30,000 characters per generation, 3 team seats, unlimited voice slots, 5 professional voice slots and a 7-day money-back guarantee on the card; everything in Plus $0.062
Max $999 $749 $8,988 billed yearly. 25,000,000 credits a month, which the card puts at up to 6,250 minutes (104 hours in the page's data); 10 team seats and 15 professional voice slots; everything in Pro $0.160
Enterprise not listed not listed Contact sales: custom volume pricing billed yearly, pay as you go with organization-level controls, zero data retention, on-premise deployment, SOC 2 compliance and larger discounts; custom SSO is coming –
API (pay as you go) not listed not listed Billed separately from the plans, with no subscription or monthly minimum: $15 per million UTF-8 bytes of text for S2.1 Pro, S2 Pro and S1 (the docs put 1 million bytes at about 180,000 English words, or about 12 hours of speech: roughly $1.25 an hour); Drama 3 at $50 per million characters, $10 until November 12, 2026, by its launch post; speech to text $0.36 an audio hour; Voice Design $0.01 a request; 5 requests at once, 15 once $100 has been paid in and 50 from $1,000 –
API S2.1 Pro Free free free The s2.1-pro-free model string: the same S2.1 Pro model at $0 under fair-use limits through November 30, 2026, with no time-to-first-audio or DPA guarantees; requests may be used to improve the model, and products earning more than $1M a year should contact Fish Audio first –

Price per minute of generated speech is our calculation: the monthly-billing price divided by what the plan includes, assuming you use all of it. It is not a measured cost.

Prices last checked: 2026-10-16 on Fish Audio's pricing page. Full breakdown: Fish Audio pricing.

Price history

No price change since we first recorded these prices on October 14, 2026. We re-check them every week and list each change here.

What hands-on testers found

We have not tested Fish Audio ourselves. These are the published tests we found that used it hands-on, newest first; follow the links for the full reviews.

Kingy AI · Curtis Pyke · updated October 13, 2026

What they tested: Fish Audio's free homepage demo on August 28, 2026, without an account, with a 200-character English script full of names, numbers and an emotional direction; the October update adds Drama 3 launch facts but no test of Drama 3, cloning or the paid API.

  • One 200-character generation with a demo voice completed and played in about 15.6 seconds.
  • Picking a second voice reset the script to Fish Audio's sample text, and the next attempt hit the anonymous demo limit and asked for a sign-in, so no controlled comparison was possible.
  • No download button appeared in the anonymous demo, and Kingy declined to score naturalness, pronunciation or emotion on unmatched scripts.

Zapier · Miguel Rebelo · updated September 8, 2026

What they tested: Fish Audio's web app among the AI voice generators he ran the same text through over three weeks, judging pacing, intonation and emotion from his ten years as an actor: the Discovery voice library and its filters, the S2 model's bracketed emotion tags and the three ways to add a voice; picked for community voice templates.

  • The community library, filtered by gender, age, use case or 48 voice qualities with a preview for most voices, is the place to look when prompting for a voice does not get you the one you want.
  • Bracketed emotion tags work, but results vary with the voice chosen, so he suggests switching voices when a take misses; voice quality across the library is inconsistent.
  • Anyone with a 10-second clip can add a voice, and fans have added TV and video game characters without the actors' consent, so picking such a voice carries copyright risk.

Feisworld Media · Fei Wu · updated May 15, 2026

What they tested: Fish Audio's voice cloning from a short clip and its text-to-speech editor with inline emotion tags and multilingual scripts, alongside ElevenLabs and Descript Overdub, as a podcaster and YouTuber (she discloses brand partnerships with ElevenLabs and other AI companies).

  • Setting up a voice clone was faster than with any other tool she had tested: 10 to 15 seconds of clear speech gave a working voice in under a minute, where ElevenLabs and Descript needed more audio and longer processing.
  • Inline emotion tags let her change the tone partway through a script instead of applying one style to the whole generation.
  • In her testing, switches between languages sounded natural rather than stitched together; she recommends it on value for creators who want cloning, multilingual output and emotion control without enterprise prices.

What users say

Test result

toolist.ai has not run its own test of Fish Audio; the findings above come from the published tests linked. See how we research and test.

Best voice tools · Fish Audio pricing

Do you make Fish Audio? Claim this listing to correct it; it's free. To see whether ChatGPT, Claude, Gemini and Perplexity recommend it, and how many readers we send you: toolist.ai for AI companies.

Sources