텍스트, 이미지, 레퍼런스 기반 워크플로우용 AI 비디오 생성기

生成視頻：圖中背景動起來，海浪動起來，紙張輕微擺動，閃電動起來。所有文字都要保持不動，文字位置也不動。視頻跟圖片比例維持不變。背景音樂要純音樂，不要有人聲。視頻尺寸是4:5。

图中女孩在商场乘电梯的视频，注意场景需要符合逻辑，声音匹配环境

理解图片是一个女生在镜子前打量自己的不同照片，根据这组照片生成一个视频，可配上合适的声音（说话或哼小曲等），需注意同时只能显示一个镜头。

Create a 15-second realistic B2B commercial video based on the reference storyboard. Subject: wholesale church chairs for large worship spaces. Gray upholstered church chairs with textured fabric cushions, padded backrests, slim black steel frames, and visible under-seat book racks. Visual style: professional, trustworthy, clean commercial product video. Warm sanctuary lighting, dark charcoal tones, warm wood interior, beige floor, premium but practical B2B advertising look. Calm, spacious, orderly atmosphere. Scene: a large modern church sanctuary facing the pulpit, filled with rows of matching gray church chairs arranged in perfect alignment. Shot sequence: 0:00-0:03 — Wide establishing shot from the back center aisle, facing the pulpit. Rows of matching church chairs fill the sanctuary symmetrically. Slow forward dolly movement. 0:03-0:06 — Medium hero shot of two gray upholstered church chairs in the foreground, with sanctuary seating behind them. Emphasize padded comfort, slim black metal frame, and commercial durability. Subtle camera push-in. 0:06-0:09 — Wide side / over-the-shoulder angle showing many rows of installed chairs. Highlight clean alignment, consistent spacing, and large-scale wholesale installation. Slow lateral camera movement. 0:09-0:12 — Close-up detail shot of the chair seat, textured gray fabric, sturdy black steel frame, and under-seat book rack. Shallow depth of field, soft highlights, durable craftsmanship. 0:12-0:15 — Final hero composition: one chair in front of a softly blurred sanctuary background, elegant lighting, clean empty space for later text overlay. Mood: calm, confident, spacious, professional, built for worship spaces. Camera: smooth dolly, slow push-in, stable commercial cinematography. Lighting: warm indoor sanctuary lighting with soft highlights on fabric and metal. Aspect ratio: 16:9. Negative prompt: no people, no clutter, no warped furniture, no distorted chair legs, no random logos, no generated text, no subtitles, no watermark, no messy background.

生成視頻：圖中人物和背景動起來，但所有字要保持不動和不變。視頻跟圖片比例維持不變。背景音樂要純音樂，不要有人聲。

生成視頻：圖中人物和背景動起來，但所有字要保持不動。視頻跟圖片比例維持不變。背景音樂要純音樂，不要有人聲。

请根据参考图的各个角度，生成一个4秒的街拍视频，场景动作切换要自然，声音匹配环境。

生成視頻：圖中人物和背景動起來，但所有字要保持不動。視頻跟圖片比例維持不變。背景音樂要純音樂，不要有人聲。

生成視頻：圖中人物和背景動起來，但所有字要保持不動。視頻跟圖片比例維持不變。背景音樂要純音樂，不要有人聲。

生成影片：圖片和背景動起來，但所有文字要保持不動。影片跟圖片維持不變。背景音樂要純音樂，不要有人聲。人物也要動起來，但要保留人物面部細節和服飾

请仔细分析参考图人脸特征，将视频的人物换成与参考图一致，注意面部表情要符合场景。电影级画质。

理解图片，生成从窗台看到“成龙”徒手在建筑物外墙向下逃逸的视频