Fish Audio
Verified BenchmarkTTS and zero-shot voice cloning with a large voice community
Quality & reliability of primary audio output.
S1/S2 TTS
Onboarding & editor ergonomics.
Free 8,000 credits/mo · Plus $15/mo ($11 annual) · Pro $100/mo · Max $999/mo
Scoring Vectors
Deterministic evaluation across the AiRecMark five-dimension rubric (v2-5dim)
# Fish Audio — official documentation
"https://fish.audio"
# Pricing source (T0): https://fish.audio/plan
# Entry tier: Free 8,000 credits/mo · Plus $15/mo ($11 annual) · Pro $100/mo · Max $999/mo (checked 2026-09-15)
# Rubric: v2-5dim · recorded features: 4Quantitative Technical Specification
Core architectural subsystems in production release v2.5
S1/S2 TTS
High-quality generation with zero-shot cloning.
Emotion control
Steer delivery and emotion tags.
Voice marketplace
Community voices and public slots.
API access
PAYG API from Plus upward.
Direct Peer Matrix: Voice Synthesis & Conversational Engines
Standardized comparative evaluation using normalized 1,000-character test payloads
| Model & Provider | AiRecMark Score | Pricing Base | Performance (rubric) | Languages | Prosody Fidelity |
|---|---|---|---|---|---|
| Fish Audio Leader | 81.7 | $15/mo | 84/100 | — | 84% |
| ElevenLabs | 95.4 | Usage API / Tiered | — | — | — |
| PlayHT | 89.2 | Freemium | — | — | — |
| Cartesia | 85.8 | $5/mo | — | — | — |
| Adobe Podcast | 85.6 | $0/mo | — | — | — |
Commercial Tiers & Compute Allocation
Transparent character pools, concurrent socket capacities, and API rate limits
Entry tier for evaluation and light individual use.
- check S1/S2 TTS
- check Emotion control
- check Voice marketplace
Free 8,000 credits/mo · Plus $15/mo ($11 annual) · Pro $100/mo · Max $999/mo
- check S1/S2 TTS
- check Emotion control
- check Voice marketplace
Dedicated infrastructure, security review, and compliance support.
- check SSO / SAML & audit controls
- check Dedicated support channel
- check Custom quota & SLA
Fish Audio stands out for 2M+ community voices across 8 languages. TTS and zero-shot voice cloning with a large voice community anchors its proposition, and AiRecMark's five-dimension audit lands it at 81.7/100 — a pragmatic default for voice-Clone Content Creators.