APEX, 조건 이동을 통한 적은 단계의 스트리밍 영상 생성
Few-Step Streaming Video Generation via Condition Shifting
- 매체
- Hugging Face
- 게시
- 중요도
- 2/5
- 화제 지수
- 19
한국어 요약
스트리밍 영상 생성의 비용을 줄이기 위해 조건 이동을 이용해 생성 분포의 속도장을 추정하는 APEX 기법을 제시했다. 180초 영상에서 기존 S2-Avatar 대비 높은 영상 지시 준수(3.856 대 3.831)와 시각적 미학성(0.641898 대 0.632606)을 달성했다. 인간 평가에서 지시 준수는 +30.0% 선호도 향상을 기록했다.
원문 발췌 · English
Few-step sampling reduces the cost of streaming video generation. Distribution Matching Distillation (DMD) supervises few-step generators with a separate fake-score network, which adds training memory and computation and must track the evolving generator. We extend APEX to streaming video by using condition shifting to estimate the velocity field of the generated distribution within the generator. The original text condition selects generation, while a shifted condition selects fake estimation using the same parameters. This removes the independent fake network while retaining S2-Avatar's frozen reference teacher, Self-Replay Forcing, and perceptual regularization. We evaluate APEX on 180-second videos using three backbone evaluations per block and a one-step spatial Refiner, with a 70-case human study and a 320-case StreamAV-Bench comparison. Compared with S2-Avatar, APEX achieves higher video instruction fulfillment on the Progressive track (3.856 versus 3.831) and higher visual aesthetics across both tracks (0.641898 versus 0.632606). Human raters favor APEX most strongly on instruction following, with +30.0% net preference, followed by consistency and video quality at +4.3% each. Motion quality favors S2-Avatar, audiovisual synchronization is predominantly tied, and usability is unchanged. Separate measurements under matched training settings show 4.4% lower peak allocated GPU memory per GPU and 9.2% shorter mean iteration time than DMD.