Nebius Token Factory
Mini toolsSoulX FlashTalk
By Soul AILab
SoulX-FlashTalk is a real-time, audio-driven avatar framework built on a 14B DiT model, designed for streaming-quality digital humans. It uses bidirectional streaming distillation to keep strong spatiotemporal coherence while cutting training cost, and achieves around 0.87 second start-up latency with 32 FPS generation on 8x H800 GPUs. The system supports long-duration talking-head animation, varied styles and even rapid speech like rap, targeting virtual anchors, digital influencers, and interactive livestream applications.
New
Audio
Gen 4
Overview
SoulX-FlashTalk is a 14B audio-driven avatar model that delivers high-fidelity lip-synced digital humans in real time, with sub-second startup and 30+ FPS streaming for live content.
About Soul AILab
Since 2016, Soul has grown into a popular AI-native platform for Generation Z. Powered by our proprietary emotional intelligence large model "Soul X" and intelligent relationship-and-content recommendation engine, we transform rich, dynamic, and exclusive public social content into immersive 24/7 experiences that deliver authentic emotional fulfillment through flow states.
Website:
soulapp.cn
Tools using SoulX FlashTalk
No tools found for this model yet.
Last updated: February 12, 2026
