task 716bc0eb-407e-4d87-9c03-56351dd41ca2 · state done · runner v7 (byte-seedance-2-omni-mini + 新 prompt + 并发)

你说原片已经很好 → 想 EN/DE/JA/KO 直接「原片→对齐→跳过 lipsync」。左边就是这个方案:各窗 v10_warp(时间轴对齐嘴动到配音节奏)裁到窗长后拼接,没做 sync-3 换嘴形。右边是当前 v7 实际产物(多一步 lipsync)。同一条配音,直接看嘴对不对得上。
@image1 = the exact first/starting frame of this clip: the video opens on precisely this image. @audio1 = the voice track of the person speaking. The person in @image1 speaks to camera and says: "Okay, so tired of those fake ads out there? This one is actually playable, I promise! Just watch these logs get perfectly stripped clean". 人物在说话时,做出一些小幅度的手势,点头,微笑,自然的身体动作。相机保持固定,人物保持看向镜头。 One single speaker only, no second person, no clone. No subtitles, no on-screen text, no watermark.
@image1 = the exact first/starting frame of this clip: the video opens on precisely this image. @audio1 = the voice track of the person speaking. The person in @image1 speaks to camera and says: "And then, oh man, they get chopped right up into tiny little pieces Oh my god, honestly the way they all pile up like that is just so oddly satisfying to watch! Wait, hold on, now the delivery trucks are totally stuck down there Come on, we really need to clear this bottleneck fast!". 人物在说话时,做出一些小幅度的手势,点头,微笑,自然的身体动作。相机保持固定,人物保持看向镜头。 One single speaker only, no second person, no clone. No subtitles, no on-screen text, no watermark.
@image1 = the exact first/starting frame of this clip: the video opens on precisely this image. @audio1 = the voice track of the person speaking. The person in @image1 speaks to camera and says: "So just hit that upgrade button and boom, instant cash rolling in! Go play Tiles Survive for free today!". 人物在说话时,做出一些小幅度的手势,点头,微笑,自然的身体动作。相机保持固定,人物保持看向镜头。 One single speaker only, no second person, no clone. No subtitles, no on-screen text, no watermark.