Cinematic Autumn Watercolor Three-Scene Video Prompt

Here's the thing.

This one is the second half of a pair. Two seconds after @Soaima_Ai posted the photo-to-watercolor autumn illustration prompt, the same account posted this, and the title line is blunt about it. Video Prompt. Not a new look, the continuation. Take the three watercolor autumn stills you just generated and turn them into one continuous clip.

The reference set is the three-panel sheet from the companion post. Top-left, the girl reaching up to touch a maple branch in a rust scarf. Top-right, the same girl in a cream knit sweater reading on a folding chair. Bottom, her riding a black bicycle with a wicker flower basket down an autumn park path.

The prompt opens with the instruction that actually does the work, the one most image-to-video prompts forget to state. Create one seamless cinematic autumn video using all three reference images in the exact same watercolor illustration style, and maintain the same girl, facial features, hairstyle, outfit, colors, proportions and artistic details throughout the entire video. Not three clips, one clip. The girl is a single character who has to survive two scene changes without her face or her sweater changing.

Scene 1. She stands beneath the autumn tree and reaches toward the falling orange leaves. Her hair and scarf move in the breeze, butterflies flutter around her, the branches sway.

Scene 2. The camera drifts through falling leaves and dissolves into her sitting under the same tree, reading. She turns the pages, butterflies circle, flowers sway, warm sunlight washes the scene.

Scene 3. Falling leaves carry the transition again, into her riding the bicycle along the park path. The flower basket sways on the handlebars, butterflies fly alongside, leaves drop from the orange trees, the vintage lamps and benches hold the dreamy mood.

The craft is in the constraints, not the scenes. Slow cinematic camera movement, gentle zoom, smooth panning, natural parallax, soft wind motion, flowing leaves, realistic subtle character movement, warm golden autumn lighting, delicate watercolor texture, 4K, smooth 24fps, vertical 9:16. Then the closing guardrail, no sudden changes, no distorted face or hands, no extra characters, no text, no watermark.

The prompt, as posted.

Video Prompt:
Create one seamless cinematic autumn video using all three reference images in the exact same watercolor illustration style. Maintain the same girl, facial features, hairstyle, outfit, colors, proportions, and artistic details throughout the entire video.
Scene 1: The girl stands beneath the autumn tree and gently reaches toward the falling orange leaves. Her hair and scarf move softly in the breeze while butterflies flutter around her. The tree branches sway naturally.
Smooth transition into Scene 2: The camera slowly moves through the falling leaves and transitions to the girl sitting peacefully under the autumn tree, reading a book. She gently turns the pages, butterflies fly around her, flowers sway, and warm sunlight softly illuminates the scene.
Smooth transition into Scene 3: The camera follows falling leaves as they transition into the girl riding her bicycle along the beautiful autumn park path. The flower basket on the bicycle gently moves, butterflies fly beside her, leaves fall from the orange trees, and the vintage street lamps and benches create a dreamy atmosphere.
Use slow cinematic camera movement, gentle zoom, smooth panning, natural parallax, soft wind motion, flowing leaves, realistic subtle character movement, warm golden autumn lighting, dreamy peaceful atmosphere, delicate watercolor texture, highly detailed animation, 4K quality, smooth 24fps, vertical 9:16.
Important: Keep all three scenes visually consistent, no sudden changes, no distorted face or hands, no extra characters, no text, no watermark, and make the transitions smooth and magical.

A small note on the original. The author is Soaima, posting as Soaima_Ai, the same account behind the photo-to-watercolor autumn illustration prompt in this library. This post carries no media of its own, it is the video half of a two-post drop, and the three reference images it points at are the three-panel sheet attached to the image prompt posted seconds earlier. If you are running this yourself, generate the three stills first, then feed all three into an image-to-video model along with this prompt. The model is not named in the post, the four-step image instructions on the companion post point at Gemini, Grok or GPT Image 2.0 for the stills, and any multi-reference image-to-video tool should take the clip.

At the time of writing, the post sits at 19 likes, 0 retweets, 31 bookmarks and 2,093 views, with no replies under it. Original thread here, x.com/Soaima_Ai/status/2096524204736737651