subject_definitions: <Subject 1> is Katia, the small… | Deepfake AI视频

作者头像
@Lostcut

生成此视频所用的文字转视频提示词

subject_definitions: <Subject 1> is Katia, the small yellow-furred cat girl in green clothing, whose design comes from the character reference sheet provided as <Picture 2>; she is rendered as a flat 2D pixel-art sprite with visible pixel edges and a limited palette. <Subject 2> is Mirmulnir, the dragon girl in green armor and a brown cloak, whose design comes from the character reference sheet provided as <Picture 3>; she is rendered in a 3D Skyrim-style game engine look with weathered textures and soft outdoor lighting. <Picture 1> is the first frame of [Shot 1], showing <Subject 2> holding <Subject 1> in front of a riverside background, with no speech bubble present. summary: [reference generation] The target video animates <Picture 1> as a single locked-off shot: <Subject 2> holds <Subject 1> beside a river, a speech bubble appears letter by letter with the line "Bakakin, can we keep her?" while <Subject 2>'s mouth moves silently, the bubble is replaced by a lower one (with speech bubble tail pointing even further down to the edge of frame, it is text coming from viewer) answering "Mirmulnir... We no longer trade Khajiit, you know?", and <Subject 2>'s expression turns to shock while <Subject 1> stays limp and blinks. retention_analysis: <Subject 1> (appears in [Shot 1]): fully_preserved - the yellow cat girl's pixel-art design, green clothing, and nervous expression are retained, with only blinking and a limp posture added. <Subject 2> (appears in [Shot 1]): fully_preserved - the dragon girl's green armor, brown cloak, and Skyrim-style 3D rendering are retained, with mouth movement and a final shocked expression added. <Picture 1> ([Shot 1] first frame): fully_preserved - the composition, background, and character placement are kept, with the speech bubble removed and only subtle environmental animation added. detailed_description: The target video is a single locked-off shot in a mixed-style look: <Subject 1> is a flat 2D pixel-art sprite, while <Subject 2> and the riverside background are rendered in a 3D Skyrim-style game engine aesthetic with weathered stone, grass, and soft daylight. [Shot 1] The shot begins exactly from <Picture 1>, with the same framing and composition: <Subject 2>, the dragon girl in green armor and a brown cloak, stands at the left of frame holding <Subject 1>, the small yellow cat girl in green clothes, cradled against her chest. The background is a riverbank with flowing water, grass, and distant trees. The camera does not move, zoom in, or zoom out at any point. The river water ripples gently and grass sways slightly, but the background layout stays unchanged. <Subject 1> hangs limply in <Subject 2>'s arms, her pixel-art expression fixed and nervous, and she blinks once or twice at irregular intervals. A speech bubble appears above <Subject 2> with no bubble in the first frame; the text inside appears letter by letter in a clean game-font style: <d>[English] Bakakin, can we keep her?</d> As the text types out, <Subject 2>'s mouth opens and closes in a talking motion, but no voice is heard. At 00:02.500, the first speech bubble disappears and a second speech bubble appears lower in the frame, its text also typing out letter by letter: <d>[English] Mirmulnir... We no longer trade Khajiit, you know.</d> At 00:04.000, the second bubble remains visible while <Subject 2>'s expression changes to shock: her eyes widen and her mouth opens slightly, her body going still. <Subject 1> keeps the same nervous pixel-art expression throughout, only blinking and hanging limp. The shot holds on this final state until the end. overall_soundscape: A continuous river-flow ambience with gentle water rushing over stones runs through the whole video, layered with distant birds chirping in the background. A soft, short text-appearance blip sound plays as each letter of the speech bubbles appears. No voices are heard. non_diegetic_music: N/A