Alternatives
Seagull Studio: a local Fish Audio alternative for Mac
Fish Audio is a cloud voice platform with monthly credit plans, a pay-as-you-go API and open-weight models you can license for self-hosting. Seagull Studio is the simpler local option: a native macOS app that clones a voice from a 10-second sample and generates speech on your own Apple Silicon Mac, offline after the model downloads, for $11.99 once with no credits to track.
What each product actually is
Fish Audio sells access to its speech models three ways: a web app on monthly plans measured in credits, a pay-as-you-go API for developers, and an enterprise self-hosting licence for its open-weight models. It is fast, multilingual, and clearly aimed at builders as much as at creators.
Seagull Studio sells one thing: a Mac app. You record or import at least 10 seconds of speech, pick one of 6 on-device models, write your script, and generate audio locally. Nothing is uploaded for synthesis, nothing is metered, and there is no account to manage — the App Store purchase is the whole transaction.
So this is not really "two clouds compared". It is a question of whether you want to rent inference from a well-run service or own a tool that runs inference on hardware you already have.
How the two compare
| Seagull Studio | Fish Audio | |
|---|---|---|
| Where synthesis happens | On your Mac | On Fish Audio servers, or self-hosted under an enterprise licence |
| Works offline | Yes, after the model is downloaded | No, unless you self-host |
| Price | $11.99 once | $0–$999 per month, or $15 per million UTF-8 bytes via API |
| Free tier | 20 successful generations | 8,000 credits per month, up to about 7 minutes, 500 characters per generation |
| Metering | None after unlock | Credits, roughly 600–625 per minute; unused credits do not roll over |
| Cloning from a short sample | Yes, from 10 seconds | Yes, instant cloning; verified professional clones need 10–180 minutes |
| Languages | Up to 31, depending on the model | 80+ with code-switching |
| API | Local MCP tools only | REST and WebSocket, Python and TypeScript SDKs |
| Open weights | No; ships selected on-device models inside the app | Yes — fish-speech, S1, S2, under a paid commercial licence |
| Platform | macOS 14+ on Apple Silicon | Any browser or backend |
| Privacy posture | Data Not Collected label; samples stay local | Cloud processing; zero data retention offered on Enterprise |
What each costs
| Plan | Monthly price | Included generation |
|---|---|---|
| Free | $0 | 8,000 credits, up to about 7 minutes; their FAQ limits free output to personal, non-commercial use |
| Plus | $15 ($66 per year) | 250,000 credits, up to about 200 minutes, 1 professional voice slot |
| Pro | $100 ($450 per year) | 2,000,000 credits, up to about 1,620 minutes, 3 seats |
| Max | $999 ($8,988 per year) | 25,000,000 credits, up to about 6,250 minutes, 10 seats |
| API | Pay as you go | $15 per million UTF-8 bytes — roughly 180,000 English words, about 12 hours of speech |
Seagull Studio is $11.99 once after 20 free generations, with restore purchases and Family Sharing. One year of Fish Audio Plus is $66 at the annual rate; Seagull Studio is less than a fifth of that, once, and then never again.
Fish Audio's API deserves credit for being honestly priced: $15 per million UTF-8 bytes with no seat fees or annual commitment, and roughly 12 hours of speech per million bytes. If you are generating audio programmatically at volume, that math can easily beat buying and running hardware. Seagull Studio is not competing for that job.
Fish Audio has open weights — why not just run them locally?
This is the fairest question on the page, because Fish Audio is closer to the local philosophy than most cloud vendors: fish-speech, S1 and S2 ship as open weights. Two things make it less simple than it sounds.
First, licensing. The open weights come with a paid commercial licence, and Fish Audio positions self-hosted deployment as an Enterprise engagement — their developer page lists $10,000 per month with an effective annual floor well into six figures. Free self-hosting means non-commercial use.
Second, operations. Weights are not a product. You need an inference server, a GPU, a model runtime, quantisation choices, an interface, and time. That is a reasonable weekend project for an ML engineer and a non-starter for a narrator who wants to record an audiobook chapter tonight.
Seagull Studio's real pitch against any open-weight option is packaging: 6 models chosen and tuned for Apple Silicon, a sample manager, generation history, language preprocessing and WAV/MP3 export, in an app you install from the App Store for $11.99.
Where Fish Audio is better
- Languages. 80+ with code-switching inside one generation, against up to 31 in Seagull Studio.
- Latency and throughput. Sub-second time-to-first-audio on their infrastructure, and concurrency that scales with spend.
- Developer surface. REST and WebSocket APIs, official SDKs, transcription and voice design endpoints.
- Verified professional cloning. Their PVC tier trains on 10–180 minutes of audio and requires the voice owner to verify by live recording, which is a genuinely good consent design.
- Any platform, plus teams. Browser access, seats on Pro and Max, and enterprise controls like zero data retention and SOC 2.
Where Seagull Studio is better
- Nothing leaves your Mac. No upload for cloning, no request per sentence, Data Not Collected on the App Store listing.
- Offline. After the model download, generation works with the network off.
- No credits. Regenerate a paragraph as many times as you like; nothing is deducted and nothing expires at the end of the month.
- One price. $11.99 once, no renewal, no plan to outgrow.
- Zero setup. No API keys, no GPU, no licence negotiation — install, record 10 seconds, generate.
Which should you pick?
Pick Seagull Studio if you work on a Mac, care that your voice stays on it, want unmetered output for narration or personal projects, and prefer buying a tool once.
Pick Fish Audio if you need 80+ languages, an API for a product, very low latency at scale, team seats, or a verified professional clone with the consent workflow that comes with it.
Pick both if that maps to your work: local for everyday generation and anything sensitive, cloud for the occasional job that needs a language or an integration Seagull Studio does not have.
Side-by-side generation examples are coming soon — the same script through Seagull Studio and Fish Audio, so you can hear the difference instead of taking anyone's word for it.
What you need to run Seagull Studio
macOS 14 or later on an Apple Silicon Mac; Intel Macs are not supported. The app is 377.5 MB and models are optional downloads from about 0.7 GB to 13 GB. Import WAV, MP3, M4A, AAC, AIFF or CAF; export WAV or MP3. Language detection is automatic, with Russian stress and Japanese kana preprocessing for cleaner reads.
Questions, answered.
Is Seagull Studio a local alternative to Fish Audio?
Yes. Fish Audio is a cloud platform with subscriptions and a pay-as-you-go API. Seagull Studio is a $11.99 macOS app that clones voices and generates speech on your own Apple Silicon Mac, offline after the model is downloaded.
Fish Audio publishes open weights — can I just run their models myself?
Technically yes, practically it depends. fish-speech, S1 and S2 ship as open weights under a paid commercial licence, and Fish Audio positions self-hosting as an Enterprise engagement starting around $10,000 per month. Running them yourself for free means a non-commercial licence, your own GPU and your own inference stack.
Which supports more languages?
Fish Audio advertises 80+ languages with instant code-switching. Seagull Studio reaches up to 31 languages, and the exact number depends on which of its 6 on-device models you choose.
Does Seagull Studio have an API for developers?
There is no HTTP API. Seagull Studio exposes local MCP tools so compatible agents running on the same Mac can list voices and models, upload a voice sample, and generate audio. Fish Audio is the better choice if you need to call a service from a backend.
Which is cheaper?
For a single person doing steady work on a Mac, Seagull Studio at $11.99 once is cheaper than any recurring plan. Fish Audio Plus is $15 per month, or $66 per year when billed annually. For high-volume programmatic generation, Fish Audio's pay-as-you-go API at $15 per million UTF-8 bytes can be more economical than buying hardware.
Do unused Fish Audio credits roll over?
No. Their pricing FAQ states monthly quotas reset each billing cycle and unused minutes do not roll over. Seagull Studio has no credits at all: after the one-time unlock, local generation is unmetered.