ElevenLabs v4
The fourth-generation text-to-speech model from ElevenLabs, supporting 90 languages. It features voice cloning from 10-second audio samples and fine-grained emotional expressiveness controls, enabling detailed modulation of synthesized speech affect and delivery.
Core Facts
Timeline (last 90 days)
ElevenLabs has released v4 and v4 Turbo, covering more than 90 languages, with Turbo inference latency of approximately 100 milliseconds, and Instant Voice Clone requiring only 10 seconds of audio to replicate a voice.
ElevenLabs v4 supports 90 languages.
The core upgrades of ElevenLabs v4 are focused on more refined expressiveness control and support for 90 languages.
ElevenLabs has launched a new v4 voice synthesis model.
There is currently a lack of sufficient third-party evaluation data on v4's audio quality details, latency performance, and horizontal comparisons with competing products.
Some voice service providers have introduced audio watermarking technology to embed invisible identifiers into synthesized audio for traceability, but whether this mechanism is applied in v4 has not been publicly disclosed.
Compressing the voice cloning sample requirement down to 10 seconds most likely leverages large-scale pre-training and few-shot learning techniques.
All Facts (6)
ElevenLabs v4 supports 90 languages.
65%VerifiedElevenLabs has launched a new v4 voice synthesis model.
65%UnverifiedThe core upgrades of ElevenLabs v4 are focused on more refined expressiveness control and support for 90 languages.
50%UnverifiedCompressing the voice cloning sample requirement down to 10 seconds most likely leverages large-scale pre-training and few-shot learning techniques.
50%UnverifiedSome voice service providers have introduced audio watermarking technology to embed invisible identifiers into synthesized audio for traceability, but whether this mechanism is applied in v4 has not been publicly disclosed.
50%UnverifiedElevenLabs has released v4 and v4 Turbo, covering more than 90 languages, with Turbo inference latency of approximately 100 milliseconds, and Instant Voice Clone requiring only 10 seconds of audio to replicate a voice.
50%