Doubao Rolls Out SeedRealtime: Free Native Full-Duplex Video Chat
xiaohu · x · 2026-08-05
ByteDance's Seed team has launched SeedRealtime, a native full-duplex audio/video model, now fully available for free in the Doubao app.
Unlike previous cascade systems or audio-only end-to-end models, the new architecture processes both visual and audio inputs within a single model. This allows it to more accurately determine when a user has finished speaking, significantly reducing interruptions. The team released 7 live demos showcasing its robust performance in noisy environments like airports and multi-person dinners.
Related event: ByteDance Launches SeedRealtime Audio-Visual Model(3 posts)→
More from Multimodal
- Running Minimax H3 on RTX 3060 Takes 1.5 Hours for a 15s Clip, Dev Seeks Optimization — SMPTHEHEDGEHOG · 2026-08-05
- Testing New Styles in Midjourney V8.2 with Exact Prompts — azed_ai · 2026-08-05
- HSWQ NVFP4 Quantization for SDXL Significantly Outperforms Native Approach — Zestyclose_Bake3680 · 2026-08-05
- Longer Prompts Degrade Image Quality: Paper Reveals Text Conditioning Scaling — rohanpaul_ai · 2026-08-05
- Creating Proper Music Videos with AI: Audio and Character Ref Workflow — Previous-Street8087 · 2026-08-05
- Minimax H3 Tested: This Node Combo Beats Spectrum in Speed — audax8177 · 2026-08-05