Fish Audio: Pricing, Voice Cloning, API and Licensing
Fish Audio is an AI voice platform for text-to-speech, voice cloning and speech-to-text, used by creators and by companies building voice products. It grew out of Fish Speech, an open-source project, and is known for expressive voices you can direct with inline tags.
The quick answer: Fish Audio is a strong, lower-cost option for narration, characters and voice apps, with a free tier and paid plans from $15 a month. Check the licensing before commercial use, because the free plan and the open models both carry limits. Its community voice library has also drawn complaints about cloning without consent.
Fish Audio at a glance
Checked against Fish Audio’s plan and developer pages on November 2, 2026.
| Fish Audio | |
|---|---|
| What it is | An AI voice platform for speech generation, cloning and transcription |
| Founders | Shijia Liao and Rissa Cao (CEO) |
| Funding | $52 million seed led by Coreline Ventures and Capital Today |
| Scale | More than 8 million users and $21 million in annual recurring revenue, per the company |
| Plans | Free, Plus at $15/month, Pro at $100/month, plus Team and Enterprise options |
| API | Pay as you go at $15 per million characters |
| Best for | Creators, developers and voice-agent builders who want expressive control |
What Fish Audio does
Fish Audio’s main products are text-to-speech and voice cloning. It offers a large library of voices, with more than 2 million listed on its site, and tools for voiceovers, audiobooks, character voices and dubbing.
Its standout feature is direction. Instead of choosing settings from a menu, you put tags right in the text, like [whispering], [excited] or [long pause]. Fish Audio says it supports more than 15,000 of these controls, and its newest featured model is called Drama 3.
TechCrunch reported that Liao, a former Nvidia researcher, trained the first model on a single GPU and open-sourced it. The Fish Speech repository has more than 31,000 stars on GitHub. Customers include HeyGen and Sanas, and Fish Audio’s developer page also names Retell.
Fish Audio reports an Audio Turing Test score of 0.515, meaning listeners in its blind evaluation could not reliably tell its S2.1 Pro model from a human. That is the company’s own test, so try your own script in your own language before you commit.
Fish Audio pricing
| Plan | Monthly | Annual billing | Generation | Limit per generation |
|---|---|---|---|---|
| Free | $0 | $0 | About 7 minutes | 500 characters |
| Plus | $15 | $11/month | About 200 minutes | 15,000 characters |
| Pro | $100 | $75/month | About 1,620 minutes | 30,000 characters |
Pro includes three team seats and five professional voice slots. Plus includes one professional slot. Fish Audio’s credits work out to roughly 600 to 625 per minute, and unused minutes do not roll over.
The API is billed separately from these plans. It costs $15 per million characters on a pay-as-you-go basis, and no subscription is needed. A free S2.1 Pro model is available through November 30, 2026, with no service guarantee, and Fish Audio says those requests may be used to improve model quality. Avoid it for anything confidential.
The licensing details to read first
Two points catch people out.
- Free plan use. Fish Audio’s FAQ says free users can only use generated audio for personal, non-commercial projects. Paying subscribers can use verified voices they own for commercial work.
- Open weights are not free for business. Fish Audio says its open models (fish-speech, S1 and S2) come with a paid commercial license. You can read and run them, but a business should expect to pay. Self-hosting is sold as an Enterprise engagement. Its newest model, S2.1 Pro, is available only through the paid API.
Voice consent is the main concern
Fish Audio built part of its library by letting people upload voices. That led to trouble. TechCrunch reported that creators alleged their voices had been uploaded without consent, and that DMCA takedowns took a long time.
CEO Rissa Cao told TechCrunch the process is now automated: a creator can submit a short sample or contract, and a voice comes down in under three minutes. That helps, but as TechCrunch notes, it doesn’t stop someone uploading an artist’s voice in the first place. The voice stays up until the artist finds out.
If you use community voices commercially, confirm you have the right to use them. Cloning a real person’s voice without permission can create legal problems for you, whatever the platform allows. For a real person’s voice, get written consent first.
How Fish Audio compares with ElevenLabs
ElevenLabs is the best-known rival, and TechCrunch also names WellSaid, Cartesia, Speechify, Async and Krisp as competitors.
Fish Audio’s appeal is price, open models and fine control of emotion through tags. ElevenLabs is the larger, more established platform, with a broad product range for enterprises. Cartesia is often considered for low-latency voice agents, and WellSaid is built around commercial and corporate narration.
The fair test is your own script. Voice quality depends on language, length and the emotion you need, so generate the same few paragraphs on each and compare the results.
Who Fish Audio suits
- Good fit: YouTubers, audiobook narrators, game developers and developers building voice agents.
- Think twice: companies with strict rules on voice rights, and anyone who needs a commercial license on a free plan.
- Worth testing: teams paying for high volume elsewhere, who should compare per-character costs on their real usage.
FAQ
Is Fish Audio free?
Yes, there is a free plan with about seven minutes of generation a month and a 500-character limit per generation. It’s for personal, non-commercial use.
How much does Fish Audio cost?
Plus costs $15 a month, or $11 a month billed annually. Pro costs $100 a month, or $75 a month billed annually. The API costs $15 per million characters.
Can I use Fish Audio commercially?
Paying subscribers can, using voices they own or have verified. The free plan is for personal projects, and the open models need a paid commercial license.
Is Fish Audio open source?
Partly. Fish Speech, S1 and S2 are available as open weights with a paid commercial license. S2.1 Pro is available through the paid API.
How much audio do you need to clone a voice?
Fish Audio’s developer page lists instant cloning from 30 seconds of audio, and some coverage cites shorter clips. More clean audio generally gives better results.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
