SepGen: One Model Generates Video with Separate Audio Stems

Tel Aviv University and Google researchers introduced SepGen, which extends a pretrained audio-video generation model to jointly output video, a mixed track, and separate waveforms per described sound source. An open-source LoRA for LTX-2.5 22B dev has also been released.

2026-10-11 ~ 2026-10-12 · 3 related posts

• Google and Tel Aviv researchers unveil SepGen, generating video with per-source stems for 4D spatial audio — YonatanBitton · 2026-10-11
• SepGen: one model generates video and per-source audio stems in a single pass — Aviad Dahan · 2026-10-12
• Open-source SepGen LoRA gives LTX-2.5 videos a separate audio track for every sound — No_Inflation9351 · 2026-10-12