A more private, cheaper alternative to Fish Audio

For developers building voice agents, conversational products, and other real-time voice applications that need multilingual speech, voice cloning, and production-scale TTS.

Open in a new tab to compare Soniox and Fish Audio on your own text.

Compare voices side by side

Quick verdict

$0.70 per generated hour

Soniox TTS v2 is one flat rate, with no premium voices or tiers billed on top.

$0.75 per generated hour

S2.1 Pro is listed at $15 per 1 million UTF-8 bytes, converted at approximately 50,000 characters per hour of speech.

Character-level timestamps

Character-level timestamps stream with the audio, while speech begins before the complete sentence is available.

Word-level timestamps

Fish Audio streams text and audio concurrently over WebSocket, with word-level timestamps available through a separate timestamp endpoint.

Zero retention by default

Soniox does not store generated audio, and never trains its models on your audio.

Retention by default

Fish Audio retains Content as long as needed for its systems to work appropriately, effectively, and efficiently. Zero Data Retention can be enabled through enterprise contracts.

Try Soniox TTS

Soniox vs Fish Audio

Price, voices, cloning, and developer tooling for each text-to-speech API.

	tts-rt-v2	s2.1-pro
Streaming audio output
Streaming text input
Multilingual voice
Voice cloning
Timestamps
Character level timestamps
Pronunciation control
Expressive control

SupportedPartialNot supported

Cost at volume

TTS costs become material when your application generates thousands of hours of speech.

$700 per month

1,000 hours of generated speech at $0.70 per hour.

$750 per month

1,000 hours of generated speech at approximately $0.75 per hour on S2.1 Pro.

Monthly TTS cost at 1,000 hours of generated speech

Soniox

tts-rt-v2

$700

Fish Audio

s2.1-pro

$750

Difference$50/month

At the published rates, the difference is $0.05 per generated hour, or $50 a month at 1,000 hours. For exact costs on your workload, use the pricing calculator.

Calculate your cost

Byte-based pricing

Fish Audio prices TTS by UTF-8 bytes of input text. Soniox prices TTS by tokens.

Billed by tokens

$4.00 per 1M input text tokens and $21.50 per 1M output audio tokens.

Billed by UTF-8 bytes

$15 per 1 million UTF-8 bytes for S2.1 Pro. A basic Latin letter uses 1 byte, while Chinese, Japanese, and Korean characters use 3 bytes.

The billing units differ. For this comparison, approximately 50,000 characters are treated as one hour of generated speech.

Concurrency

Built for high concurrency

Soniox TTS runs on production infrastructure built for high-concurrency workloads, with 99.9% uptime and regional deployment.

Published concurrency tiers

Fish Audio allows 5 concurrent requests below the first paid threshold, 15 after the next threshold, and 50 at high volume; enterprise limits are custom.

Voice cloning

Clone from a short reference clip

Upload up to 20 seconds of audio and get a voice ID you use like any built-in voice. Soniox removes background noise and echo, and cloned voices work across all 60+ supported languages.

Persistent or instant cloning

Fish Audio creates persistent voices with a reference ID, while instant cloning passes reference audio inline on each request. It recommends at least 10 seconds per clip and says longer clear, single-speaker recordings improve fidelity.

Can I bring my existing Fish Audio voice?

Not as an existing Fish Audio voice ID. Soniox creates its own voice from a reference clip. If you have the rights or permission to use the underlying voice recording, you can use a reference clip to create a Soniox voice.

Explore Soniox Voice Cloning

Real-time voice generation

WebSocket streaming and REST

Speech begins before the complete sentence is available. Character-level timestamps stream with the audio, so your app knows exactly what was spoken when a user interrupts.

WebSocket streaming and REST

Fish Audio accepts incremental text chunks and returns audio chunks concurrently over WebSocket. Word-level timestamps are available through separate timestamp endpoints.

Switch from Fish Audio in minutes

Soniox provides REST and real-time TTS APIs, plus official SDKs for Python, Node.js, Web, React, and React Native.

import requests

response = requests.post(
    "https://tts-rt.soniox.com/tts",
    headers={
        "Authorization": f"Bearer {SONIOX_API_KEY}",
    },
    json={
        "model": "tts-rt-v2",
        "language": "en",
        "voice": "Adrian",
        "audio_format": "mp3",
        "text": "Hello from Soniox.",
    },
)

audio = response.content

import requests

response = requests.post(
    "https://api.fish.audio/v1/tts",
    headers={
        "Authorization": f"Bearer {FISH_API_KEY}",
        "model": "s2.1-pro",
    },
    json={
        "text": "Hello from Fish Audio.",
        "reference_id": "YOUR_VOICE_ID",
        "format": "mp3",
    },
)

audio = response.content

A Fish Audio request that sends text to a voice becomes a Soniox request with a model, language, voice, and text. A cloned voice is referenced by its voice ID.

Start your migration

Privacy and data handling

Zero retention by default

Soniox does not store generated audio, and never trains its models on your audio. Speech data is processed within your selected region, with regional deployment for data residency requirements.

Retention and enterprise controls

Fish Audio retains Content as long as needed for its systems, and its terms allow Usage Data and Content to be used to develop, train, or enhance AI or machine learning models. Zero Data Retention requires an enterprise contract, and data stays in the United States by default.

Fish Audio alternative FAQ

How much does Soniox TTS cost compared with Fish Audio?

Soniox TTS v2 is listed at $0.70 per generated hour. Fish Audio S2.1 Pro is listed at $15 per 1 million UTF-8 bytes, approximately $0.75 per generated hour.

Does Soniox support voice cloning?

Yes. Soniox can create a custom voice from a short reference clip. The cloned voice can then be used by voice ID across Soniox’s supported languages.

Can I keep my existing Fish Audio voices?

You cannot transfer a Fish Audio voice ID directly to Soniox. If you have the rights or permission to use the source recording, you can use a reference clip to create a Soniox voice.

Does Soniox support real-time TTS?

Yes. Soniox streams speech through its real-time API and can begin generating speech while text is still arriving.

How many languages does Soniox TTS support?

Soniox TTS supports 60+ languages, and every voice can speak every supported language.

Does Soniox have an API and SDKs?

Yes. Soniox provides REST and real-time APIs and official SDKs for Python, Node.js, Web, React, and React Native.

How does Soniox pricing work?

Soniox TTS pricing is $4.00 per 1M input text tokens and $21.50 per 1M output audio tokens. That comes to about $0.70 per generated hour.

Ready to get started?

Create an account instantly, or contact us to design a custom package for your business.

Build with API

Documentation

Get up and running in minutes and spend your time building, not wrestling with the API.

Explore docs

See what you’ll pay

Pay only for what you use with our flexible pricing. Built to scale with you.

Pricing details