AI BriefingElevenLabs Unveils 'Flash', an Ultra-Low-Latency Voice Generation Model
Key point
ElevenLabs has launched Flash, an ultra-low-latency voice generation model optimized for conversational voice agents.
Details
ElevenLabs has unveiled Flash, a new model for conversational voice agents. This model delivers an overwhelming low-latency performance of 75ms, excluding application and network latency.
Flash is available immediately on the Conversational AI platform, and can also be built directly via the API using the eleven_flash_v2 and eleven_flash_v2_5 model IDs.
Detailed specifications for each model version are as follows:
- Flash v2: English-only support
- Flash v2.5: Support for 32 languages
- Cost: Consumes 1 credit per 2 characters
While Flash's quality and depth of emotional expression may be slightly lower compared to the Turbo model, its latency is much shorter. In blind tests by human evaluators, it outperformed similar ultra-low-latency models, proving it has best-in-class speed and quality.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
StepAudio 3 Gen Achieves SOTA in TTS and Multimodal Audio Generation via RVQ-Based Discrete GenerationUMG Partners with ElevenLabs to Launch AI Music Platform