Eleven v3

ElevenLabs

Eleven v3 is a text-to-speech model from ElevenLabs, generally available since February 2, 2026, supporting 70+ languages with inline audio tags for emotional delivery. ElevenLabs now describes it as its previous-generation speech synthesis model, following the launch of Eleven v4 and v4 Turbo on 28 September 2026. ElevenLabs also offers an Eleven v3 Conversational variant for real-time use, and prices the model through its subscription plans.

Full Specifications

All specsBest valueCheapestHeavy workloadMax capabilityDifferences only

	Eleven v3ElevenLabs Included in ElevenLabs subscription plans - see vendor pricing
Identity
Developer	ElevenLabs
Released	Feb 2026
Status	GA - "Eleven v3 is now generally available across all platforms" since February 2, 2026, out of the prior Alpha label
Licence	Proprietary
Self-hostable	No
Cost
Price per minute	API pay-as-you-go: ~$0.10/minute for Multilingual v2 and v3 (billed at $0.10 per 1,000 characters; ElevenLabs' own API pricing page groups v2 and v3 at the same rate, no v3 premium); ~$0.05/minute for Flash v2.5/Turbo v2.5. Text-to-Speech is credit-metered at 1 credit per character (not natively per-minute) - subscription plans instead bill in monthly Credits rather than a flat per-minute rate (see Cheapest plan)
Cheapest plan	Starter $6/mo (30,000 credits/mo) is the cheapest paid subscription; Free ($0/mo, 10,000 credits/mo) also has Text to Speech access but excludes Commercial License and Instant Voice Cloning, both listed as included only from Starter up in ElevenLabs' own plan-comparison table (see Free tier); Professional Voice Cloning requires Creator ($22/mo, 121,000 credits/mo) or above, though ElevenLabs' own docs do not list Eleven v3 among Professional Voice Cloning's trained models - see Voice cloning
Free tier	Yes - Free plan ($0/mo, 10,000 credits/mo) includes Text to Speech, but ElevenLabs' own plan-comparison table lists "Commercial License" and "Instant Voice Cloning" as included only from the Starter plan up, not on Free
Capacity
Max length	5,000 characters per request (~5 minutes of audio, per ElevenLabs' own model docs) - shorter than Multilingual v2's 10,000-character limit and Flash v2.5's 40,000-character limit, a tradeoff for v3's emotional-delivery focus
Sample rate	Up to 48kHz - the output_format API parameter offers PCM/WAV/Opus at 48000 Hz alongside lower rates (8000-44100 Hz) and MP3 up to 44.1kHz; default output is mp3_44100_128 (44.1kHz/128kbps). Higher-bitrate MP3 formats require Creator tier+; PCM/WAV formats require Pro tier+
Output formats	MP3 (22050-44100 Hz, 32-192kbps), Opus (48000 Hz, 32-192kbps), PCM (8000-48000 Hz), WAV (8000-48000 Hz), A-law and mu-law (8000 Hz) - selectable via the output_format API parameter; default mp3_44100_128
Capability
Voice cloning	Yes - Instant Voice Cloning (zero-shot from 1-2 min of audio, no plan restriction, usable with any model including v3) and Professional Voice Cloning (dedicated fine-tuned model, Creator plan+, 30-180 min of audio, 3-6+ hr training); however ElevenLabs' own Professional Voice Cloning docs list only Flash v2.5, Turbo v2.5 and Multilingual v2 as the models a PVC automatically trains on - Eleven v3 is not among them
Languages	70+ languages (99 listed by name/ISO code on ElevenLabs' own v3 product page), from Afrikaans and Arabic through Vietnamese and Welsh
Stem separation	Not offered - Eleven v3 is a text-to-speech/dialogue model; ElevenLabs' product lineup includes a separate "Voice Isolator" (removes background noise from a single vocal track) but no feature that splits a music mix into individual instrument/vocal stems
Rights
Commercial licence	Paid plans only - "Commercial License" is absent from the Free plan and listed as included from Starter ($6/mo) upward in ElevenLabs' own plan-comparison table
Access
API	Yes - "Public API for Eleven v3 is now available" per ElevenLabs' own v3 product page, with official Python (elevenlabs) and Node.js (@elevenlabs/elevenlabs-js) client libraries
Chat app	Yes - ElevenLabs' own web Speech Synthesis studio (elevenlabs.io/text-to-speech) plus the ElevenReader companion app (web, iOS, Android) for text-to-speech playback
Cloud marketplaces	Yes - Google Cloud Marketplace, a first-party ElevenLabs listing per ElevenLabs' own announcement ("ElevenLabs' advanced voice AI models are now available on Google Cloud Marketplace"); not confirmed as a first-party listing on AWS Marketplace (only third-party solutions built using ElevenLabs found there) or Microsoft/Azure Marketplace (the one listing found there is published by a third party, not ElevenLabs)
Fine-tuning	Not offered for Eleven v3 specifically - Professional Voice Cloning is ElevenLabs' fine-tuning feature, but its own docs state PVCs "automatically train on the Flash v2.5, Turbo v2.5 and Multilingual v2 models" only; Eleven v3 is not listed among the models a Professional Voice Clone trains on