Open SourceStar on GitHub, add a model

Which open-source TTS should I try?

Start with the three picks below. Hear them. Pick one. We’ll get into the rest later.

🥇
Best overall
Kokoro-82M

The current default recommendation.

Excellent quality. Tiny model. Runs almost anywhere.

Voice
Read more
🥈
Runner-up
Orpheus TTS

Most natural-sounding conversations.

Llama-3 based. Empathetic. Built for interactive apps.

Voice
Read more
🥉
Third place
Chatterbox

Best voice cloning, MIT-licensed.

ElevenLabs-tier quality from a few seconds of reference audio.

VoiceDefault
Read more
Side-by-side

Hear the leaders back-to-back

Same three scripts, three top models, one screen. The fastest way to decide what you actually want.

Kokoro-82MOrpheus TTSChatterbox
Compare these three
The three scripts (every voice reads these)
Neutral
The quick brown fox jumps over the lazy dog near the riverbank.
Baseline naturalness — no emotion, no tricky words
Emotional
I can't believe you actually did it. This is incredible!
Prosody and excitement — tests expressiveness
Numbers & Dates
On March 14th, 2025, the team raised $4.2 million at a 38% margin.
Number, date, and symbol pronunciation — the common failure mode
Browse by need

Not sure what you need yet?

Each shelf below is hand-curated. Scroll horizontally to see all picks. Add anything to the comparison bar to hear them side-by-side.

Best for voice cloning

Clone a voice from a short reference clip.

Chatterbox

Default

ElevenLabs-tier quality, MIT licensed, voice cloning from a clip

CommercialCloningStreaming500M· 8GB

XTTS v2

Reference clone

Most-downloaded TTS on HF, 6-second voice cloning

CloningStreaming750M· 6GB

Fish Speech

Reference clone

Open-source TTS that benchmarks above closed-source

CommercialCloningStreaming0.5B / 4B· 12GB

Bark

Generates laughter, sighs, and music — the chaotic one

CommercialCloning1.5B· 12GB

F5-TTS

Reference clone

Flow-matching TTS, fast zero-shot voice cloning

CommercialCloning335M· 8GB

Dia

Ultra-realistic two-speaker dialogue, sounds like a podcast

CommercialCloningStreaming1.6B· 10GB

Orpheus TTS

Llama-3 based, empathetic, low-latency for interactive apps

CommercialCloningStreaming3B· 12GB

Best for real-time

Stream tokens fast enough for voice agents.

Kokoro-82M

82M params, Apache 2.0, runs on a potato

CommercialStreaming82M· 1GB

Parler-TTS

Describe the voice in natural language ('soft female, fast, clear')

CommercialStreaming880M· 6GB

Fish Speech

Reference clone

Open-source TTS that benchmarks above closed-source

CommercialCloningStreaming0.5B / 4B· 12GB

Orpheus TTS

Llama-3 based, empathetic, low-latency for interactive apps

CommercialCloningStreaming3B· 12GB

Chatterbox

Default

ElevenLabs-tier quality, MIT licensed, voice cloning from a clip

CommercialCloningStreaming500M· 8GB

XTTS v2

Reference clone

Most-downloaded TTS on HF, 6-second voice cloning

CloningStreaming750M· 6GB

Dia

Ultra-realistic two-speaker dialogue, sounds like a podcast

CommercialCloningStreaming1.6B· 10GB

Best for multilingual

Five or more languages in one model.

Chatterbox

Default

ElevenLabs-tier quality, MIT licensed, voice cloning from a clip

CommercialCloningStreaming500M· 8GB

XTTS v2

Reference clone

Most-downloaded TTS on HF, 6-second voice cloning

CloningStreaming750M· 6GB

Bark

Generates laughter, sighs, and music — the chaotic one

CommercialCloning1.5B· 12GB

Orpheus TTS

Llama-3 based, empathetic, low-latency for interactive apps

CommercialCloningStreaming3B· 12GB

Kokoro-82M

82M params, Apache 2.0, runs on a potato

CommercialStreaming82M· 1GB

Fish Speech

Reference clone

Open-source TTS that benchmarks above closed-source

CommercialCloningStreaming0.5B / 4B· 12GB

Best for edge deployment

Under 200M parameters — embed anywhere.

Kokoro-82M

82M params, Apache 2.0, runs on a potato

CommercialStreaming82M· 1GB

Best for emotion + flavor

Long-form, dialogue, and expressive models.

Bark

Generates laughter, sighs, and music — the chaotic one

CommercialCloning1.5B· 12GB

Dia

Ultra-realistic two-speaker dialogue, sounds like a podcast

CommercialCloningStreaming1.6B· 10GB

Want the full directory?

All 26 models with full filters and sort. 46 voices total. For when you're ready to do your own research.

Browse all 26 models

Missing a model? Add it.

OpenSpeech is community-maintained. Adding a model is a single PR: edit one JSON file, drop in three audio samples, open the pull request. Models must be genuinely open-source.