Text-to-speech · registry
Reported MOS is metadata, not a measured leaderboard.
This page keeps vendor, paper, community, and CodeSOTA-measured rows visible in one registry. It avoids assigning a strict rank to reported MOS and flags sub-0.1 MOS gaps as noise.
| Model | Type | Evidence tier | MOS | MOS interpretation | Source | Deployment | License | Capabilities |
|---|---|---|---|---|---|---|---|---|
Gradium TTS | API | codesota measured | 4.4 | CodeSOTA measured | api | provider terms | streaming · cloning · voice design | |
Kokoro v1.0 | open | codesota measured | 4.5 | CodeSOTA measured | local, edge | Apache-2.0 | streaming | |
Fish Audio S2 Pro | open | paper reported | 4.6 | paper reported | local, edge, browser | fish-audio-research-license | streaming · cloning · voice design · code-switching | |
Supertonic 3 | open | community reported | 4.2 | community reported | local, edge, browser | OpenRAIL-M | cloning · voice design · code-switching | |
Sesame CSM | open | community reported | 4.7 | within MOS noise | community reported | local | Apache-2.0 | voice design |
Gemini 2.5 Pro TTS | API | vendor reported | 4.7 | within MOS noise | vendor reported | api | provider terms | streaming · voice design · code-switching |
Cartesia Sonic 2 | API | vendor reported | 4.7 | within MOS noise | vendor reported | api | provider terms | streaming · cloning · voice design |
ElevenLabs Flash v2.5 | API | vendor reported | 4.6 | vendor reported | api | provider terms | streaming · cloning · voice design · code-switching | |
PlayHT 3.0 | API | vendor reported | 4.6 | vendor reported | api | provider terms | streaming · cloning · voice design | |
Orpheus TTS | open | community reported | 4.6 | community reported | local | Apache-2.0 | cloning · voice design | |
Gemini 2.5 Flash TTS | API | vendor reported | 4.5 | vendor reported | api | provider terms | streaming · voice design · code-switching | |
Google Chirp 3 HD | API | vendor reported | 4.4 | vendor reported | api | provider terms | streaming · cloning · voice design · code-switching | |
Fish Speech 1.5 | open | community reported | 4.4 | community reported | local | Apache-2.0 | cloning · code-switching | |
Dia 1.6B | open | community reported | 4.3 | community reported | local | Apache-2.0 | cloning · voice design | |
Spark-TTS | open | community reported | 4.3 | community reported | local | Apache-2.0 | cloning · voice design · code-switching | |
Parler-TTS | open | paper reported | 4.1 | paper reported | local | Apache-2.0 | voice design | |
ElevenLabs Turbo v2.5 | API | vendor reported | 4.8 | within MOS noise | vendor reported | api | provider terms | streaming · cloning · voice design · code-switching |
XTTS v2 | open | paper reported | 4.5 | paper reported | local | CPML | cloning · code-switching | |
F5-TTS | open | paper reported | 4.4 | paper reported | local | MIT | cloning · code-switching | |
OpenAI TTS HD | API | vendor reported | 4.7 | within MOS noise | vendor reported | api | provider terms | streaming |
Piper | open | community reported | 3.6 | community reported | local, edge | MIT | streaming |