Fish S2 Pro
Preview$0.015
$15.00
2026-03-09
- Positioning
- Flagship, as the vendor describes it. The vendor has a newer generation.Source
- Price
- Level 2 of 4 among audio models ($ is the cheapest), by price per 1,000 characters: $0.01 to under $0.02.
- Vendor-stated speed
- Not shown: the vendor doesn't compare this model's speed.
Three separate signals, not a quality score.
Fish Audio S2 Pro is Fish Audio's openly released text-to-speech model, trained on more than 10 million hours of audio across 80+ languages. Free-form inline tags such as [whisper] or [excited] steer emotion and prosody word by word, and it handles multi-speaker, multi-turn dialogue. Fish Audio now lists it as its previous-generation S2 model, succeeded by S2.1-Pro.
Pricing
Text-to-speech (Fish Audio), billed at USD 15 per 1M characters.
Where it fits
- Expressive narration and character voices steered with inline emotion tags
- Multi-speaker dialogue such as podcast-style or drama scripts
- Multilingual voiceover across Chinese, English, Japanese and dozens of other languages
- Budgets that track script length, since ATP bills per input character
- Speech recognition or audio understanding — this id is text-to-speech only
- Commercial self-hosting of the open weights — they ship under a research, non-commercial license
Benchmarks
- Seed-TTS Eval WER (English)0.99% WER
lower is better; 'best overall' per vendor
Fish Audio - Seed-TTS Eval WER (Chinese)0.54% WER
lower is better; 'best overall' per vendor
Fish Audio - EmergentTTS-Eval win rate81.88%
win rate vs gpt-4o-mini-tts baseline
Fish Audio
| Benchmark | Score | Source |
|---|---|---|
| Seed-TTS Eval WER (English)lower is better; 'best overall' per vendor | 0.99% WER | Fish Audio |
| Seed-TTS Eval WER (Chinese)lower is better; 'best overall' per vendor | 0.54% WER | Fish Audio |
| EmergentTTS-Eval win ratewin rate vs gpt-4o-mini-tts baseline | 81.88% | Fish Audio |
Scores as published by each vendor, not measured by ATP. Test conditions vary, so they are not directly comparable across vendors.
Same controls on every model: project keys, per-project model access, credit budgets, and a log of every request.Set up your organization
Cost estimator
- Characters$7.50
$7.50 USD / month · 750 credits / month
Quickstart
curl -s https://api.atptoken.ai/v1/models \
-H "Authorization: Bearer atp-•••"
# data[].id: "fish-s2-pro"FAQ
- What can I send Fish S2 Pro, and what do I get back?
- Fish S2 Pro is a Fish Audio text-to-speech model. Send the text to speak and you get back speech audio.
- How is Fish S2 Pro priced?
- You pay per 1,000 characters of text you send. You only pay when a request succeeds. $1 = 100 credits. See pricing
- How do I call Fish S2 Pro on ATP?
- Create a project API key. Check that "fish-s2-pro" shows up with GET https://api.atptoken.ai/v1/models, then follow the speech generation docs with model set to that id.
Sources
Start with $5. Credits never expire.
Create a project key, turn this model on, and pay only for what you use.