Modality console · /voice/

Voice

TTS, voice cloning and speech. Open axes, cited sources, trust flags. Quality is metered only from an independent benchmark — no benchmark, no meter. No single composite score.

13 tools classes A·B·C data as of 2026-07-04 refresh · monthly

segment Best free Real free output — not a trial. segment Best for API Public self-serve API, ordered by price. segment Open-weights Open model licence — self-hostable. segment Voice cloning Clone a voice from a sample.

Filter · in-page 13 shown

Cartesia (Sonic) Cartesia B live

API & price: public · credit-based (~$50/1M chars)
Voices: not published
Voice cloning: yes · instant
Languages: 42
Weights: closed
Free tier: limited

Quality 1209Artificial Analysis Speech Arena (Sonic 3.5, #2) ↗

Best for APIVoice cloningBest free ✓ 2026-07-04

Chatterbox Resemble AI A live

API & price: self-host only (open weights)
Voices: any (design)
Voice cloning: yes · instant
Languages: 23
Weights: open · MIT
Free tier: watermarked

Quality 1479TTS Arena V2 ↗

Open-weightsVoice cloningBest free ✓ 2026-07-04

Deepgram Aura Deepgram A live

API & price: public · $30/1M chars (Aura-2); $15/1M (Aura-1)
Voices: ~90 (Aura-2)
Voice cloning: no
Languages: 7
Weights: closed
Free tier: limited

Quality no independent measure

Best for APIBest free ✓ 2026-07-04

ElevenLabs ElevenLabs A live

API & price: public · $100/1M chars (v2/v3), $50/1M (Flash/Turbo)
Voices: 10,000+ (community library)
Voice cloning: yes · instant
Languages: 32
Weights: closed
Free tier: limited

Quality 1172Artificial Analysis Speech Arena (Eleven v3) ↗

Best for APIVoice cloningBest free ✓ 2026-07-04

Fish Audio (OpenAudio) Fish Audio B live

API & price: public · ~$15/1M chars (PAYG)
Voices: clone/slot-driven library
Voice cloning: yes · instant
Languages: 80+
Weights: open · Fish Audio Research License
Free tier: open self-host

Quality 1146Artificial Analysis Speech Arena (S2.1 Pro) ↗

Best for APIVoice cloningBest freeOpen-weights ✓ 2026-07-04

Hume AI (Octave) Hume AI A live

API & price: public · ~$0.05–0.15/1K chars (tiered)
Voices: any (design)
Voice cloning: yes · instant
Languages: 11
Weights: closed
Free tier: limited

Quality 1060Artificial Analysis Speech Arena (Octave 2) ↗

Best for APIVoice cloningBest free ✓ 2026-07-04

Kokoro hexgrad A live

API & price: self-host only (no official vendor API)
Voices: 54
Voice cloning: no
Languages: 8
Weights: open · Apache-2.0
Free tier: open self-host

Quality 1059Artificial Analysis Speech Arena (Kokoro 82M) ↗

Open-weightsBest free ✓ 2026-07-04

Microsoft Azure TTS Microsoft B live

API & price: public · ~$16/1M chars (Neural); ~$22/1M (Neural HD)
Voices: 600+
Voice cloning: yes · pro only
Languages: 140+
Weights: closed
Free tier: limited (500K chars/mo free)

Quality 1125Artificial Analysis Speech Arena (Azure HD 2.5) ↗

Best for APIVoice cloningBest free ✓ 2026-07-04

MiniMax Speech MiniMax B live

API & price: public · ~$100/1M chars (Speech-02-HD)
Voices: 300+
Voice cloning: yes · instant
Languages: 30+
Weights: closed
Free tier: limited

Quality 1184Artificial Analysis Speech Arena (Speech 2.8 HD) ↗

Best for APIVoice cloningBest free ✓ 2026-07-04

OpenAI TTS / Realtime OpenAI A live

API & price: public · tts-1 $15/1M chars; Realtime token-priced
Voices: 11
Voice cloning: no
Languages: ~57 (Whisper set)
Weights: closed
Free tier: no

Quality 1100Artificial Analysis Speech Arena (TTS-1 HD) ↗

Best for API ✓ 2026-07-04

PlayHT / PlayAI PlayAI (acquihired by Meta) C defunct

API & price: discontinued (was ~$15–100/1M chars)
Voices: 900+ (historical)
Voice cloning: yes · instant (historical)
Languages: 142 (historical, claimed)
Weights: closed
Free tier: no

Quality no independent measure

✓ 2026-07-04

Resemble AI Resemble AI B live

API & price: public · pay-as-you-go (~$0.0005/sec, disputed)
Voices: not published
Voice cloning: yes · instant
Languages: 23
Weights: closed
Free tier: limited

Quality no independent measure

Best for APIVoice cloningBest free ✓ 2026-07-04

Sesame CSM Sesame A research

API & price: self-host only (open weights)
Voices: any (design)
Voice cloning: no (contextual matching, not packaged clone)
Languages: 1 (English)
Weights: open · Apache-2.0
Free tier: open self-host

Quality no independent measure

Open-weightsBest free ✓ 2026-07-04