Models

Voice models compared

Vendor-stated figures only, checked October 14, 2026. A dash means the vendor doesn't publish it.

Find my voice model →

ElevenLabs Flash v2.5

ElevenLabs

TTS

Latency: ~75 ms (model inference)
Price: $0.04 / 1K chars
Languages: 32
Cloning: Yes

▶ Demo ↗Source

ElevenLabs v4 Turbo

ElevenLabs

TTS

Latency: ~100 ms median (model inference)
Price: $0.04 / 1K chars
Languages: 90+
Cloning: Yes

▶ Demo ↗Source

Cartesia Sonic 3.6

Cartesia

TTS

Latency: —
Price: Free tier; from $5/mo (100K credits)
Languages: 44
Cloning: Yes (instant, paid plans)

▶ Demo ↗Source

Rime

TTS

Latency: 37 ms P50 time-to-first-audio
Price: $0.03 / 1K chars (~$0.03/min)
Languages: 4 (en, fr, de, es)
Cloning: Enterprise

▶ Demo ↗Source

Rime

TTS

Latency: 96 ms P50 time-to-first-audio
Price: $0.05 / 1K chars (~$0.05/min)
Languages: 8
Cloning: Enterprise

▶ Demo ↗Source

Deepgram Aura-2

Deepgram

TTS

Latency: —
Price: $0.030 / 1K chars
Languages: 7 (en, es, de, fr, nl, it, ja)
Cloning: —

▶ Demo ↗Source

Deepgram

STT

Latency: Built-in turn detection
Price: $0.0065/min (English)
Languages: English or multilingual
Cloning: —

ElevenLabs Scribe v2 Realtime

ElevenLabs

STT

Latency: ~150 ms partials
Price: $0.39 / hr
Languages: 90+
Cloning: —

OpenAI gpt-realtime-2.1

OpenAI

Speech-to-speech

Latency: —
Price: Audio $32 in / $64 out per 1M tokens
Languages: —
Cloning: —

▶ Demo ↗Source

OpenAI gpt-realtime-2.1-mini

OpenAI

Speech-to-speech

Latency: —
Price: Audio $10 in / $20 out per 1M tokens
Languages: —
Cloning: —

▶ Demo ↗Source

Model	Type	Latency (vendor claim)	Languages	Price	Voice cloning	Hear it
ElevenLabs Flash v2.5ElevenLabs	TTS	~75 ms (model inference)	32	$0.04 / 1K chars	Yes	▶Demo ↗	Source
ElevenLabs v4 TurboElevenLabs	TTS	~100 ms median (model inference)	90+	$0.04 / 1K chars	Yes	▶Demo ↗	Source
Cartesia Sonic 3.6Cartesia	TTS	—	44	Free tier; from $5/mo (100K credits)	Yes (instant, paid plans)	▶Demo ↗	Source
Rime Mist v3Rime	TTS	37 ms P50 time-to-first-audio	4 (en, fr, de, es)	$0.03 / 1K chars (~$0.03/min)	Enterprise	▶Demo ↗	Source
Rime CodaRime	TTS	96 ms P50 time-to-first-audio	8	$0.05 / 1K chars (~$0.05/min)	Enterprise	▶Demo ↗	Source
Deepgram Aura-2Deepgram	TTS	—	7 (en, es, de, fr, nl, it, ja)	$0.030 / 1K chars	—	▶Demo ↗	Source
Deepgram FluxDeepgram	STT	Built-in turn detection	English or multilingual	$0.0065/min (English)	—	—	Source
ElevenLabs Scribe v2 RealtimeElevenLabs	STT	~150 ms partials	90+	$0.39 / hr	—	—	Source
OpenAI gpt-realtime-2.1OpenAI	Speech-to-speech	—	—	Audio $32 in / $64 out per 1M tokens	—	▶Demo ↗	Source
OpenAI gpt-realtime-2.1-miniOpenAI	Speech-to-speech	—	—	Audio $10 in / $20 out per 1M tokens	—	▶Demo ↗	Source

ElevenLabs latency excludes application and network time. Samples open on each vendor’s own demo page; we don’t rehost audio. PlayHT omitted: no verifiable current pricing.