by Jaime Yao
Flick Residency Films Premiere Officially. See behind the scenes
Watch on FlickEach shot was first blocked out in Blender to fix the camera. Jaime's blog post explains that the Blender screenshots looked synthetic when turned straight into images, so he reduced them to simple sketches first. About 70 sketches were uploaded, and 229 image prompts open the same way: the sketch decides the composition and the camera.
Use the provided sketch as the only composition and camera reference
Almost every image prompt (800 of 831 kept images) starts by listing its references and what each one is for: the sketch for framing, a three-view sheet for each person, a detail photo for the bedding. Naming the job of each image keeps the model from mixing them up.
use the provided three-view girl reference as the only girl reference
The knight started as an uploaded three-view sheet. Nano Banana Pro then added his scroll case, grappling hook and dagger, and that version became the most used image in the workflow, a reference in 167 generations. The girl and her father have three-view sheets of their own.
Add three accessories to the existing knight turnaround sheet
The cardboard box got the same treatment as the characters: a sheet with every side, notes on size and material, the castle print and the knight drawn in for scale. A lidless version of the sheet was then used in 20 generations.
Change this box reference sheet into a lidless version.
The knight drawing on paper was generated from the knight photo, with a prompt that asks for what a child would actually draw: a big head, stiff limbs, simple shapes and uneven lines. That drawing was reused in 29 generations.
Draw it as a crayon drawing made sincerely from memory by an 8- or 9-year-old child.
To show how small the knight is, the camera sits near the floor and looks up at the furniture around him. 122 prompts ask for a low angle or a camera near the floor.
The camera is very low, near the floor, looking upward.
The frames were animated mostly with Kling O3 Pro (156 clips) and Seedance 2.0 (49). 191 of the 206 kept video prompts begin with the camera, handheld or locked, then describe the action one line at a time, and 163 ask for no background music.
Locked camera. A small sound comes from inside first. A tiny little knight struggles to climb up from inside onto the windowsill.
Handheld camera style. Keep the focus locked on the foreground man the whole time. The foreground man stays sharp the whole time. The little girl in the background stays out of focus the whole time. Starting from the doorway, the little girl keeps walking forward while looking ahead, getting closer and closer to the foreground man. She continues moving closer throughout the shot, not just taking one small step. The man hears her and slowly looks up at her. In a soft, gentle, reassuring voice, he says, “Lydia, would you like to tell me his story?” The little girl replies with a slightly surprised, delighted voice, “All of it?” The man smiles and answers, “Yes, all of it.” Do not rack focus. Do not cut to another shot.
Locked camera. A small sound comes from inside first. A tiny little knight struggles to climb up from inside onto the windowsill. He has an engraved scroll case strapped diagonally across his back, with a grappling hook and a sword at his waist. The top of the scroll case has a clearly visible closed metal end cap, short and rounded, and it must clearly read as capped and sealed rather than open. After climbing onto the sill, he braces himself with one hand on the window frame, pauses, and catches his breath. The action should feel slightly effortful. The camera remains completely still throughout. No background music.
Locked camera. Keep the focus on the foreground windowsill, the little knight, and the drawing paper the whole time. The blurred father and little girl in the background keep sitting on the bed and looking at the drawing, with only very subtle natural movement. The little knight gently taps the drawing beneath him with both hands. The trees in the drawing gradually loosen into colored-pencil dust and soft pigment particles, then slowly rise into the air. The colored dust gathers again on the paper and gradually transforms into several highly realistic miniature trees. Quiet indoor ambience. No background music.
Handheld camera. The camera pulls back farther. The little girl stays in the doorway, looking into the room. As the camera pulls back, a man in a blue-gray sweater appears in the foreground. He is sitting at the foot of the bed. His legs resting on the floor are clearly visible. He keeps his head down and looks at several child’s drawings resting on his knees. His attention stays on the drawings on his knees and he does not look toward the doorway. By the end of the shot, both the little girl in the doorway and the man sitting at the foot of the bed are visible in the same frame. No background music.
Locked camera.The blurred father and little girl in the background keep sitting on the bed and looking at the drawing, with only very subtle natural movement. The knight drawing on the windowsill stays in place the whole time. The little knight inside the drawing slowly breaks apart into colored-pencil dust and soft pigment particles, rising gently from the paper. The colored dust gathers in the air and gradually forms a small three-dimensional figurine knight. At the end, the figurine knight appears sitting on the edge of the bed. No background music.
Use the provided box reference as the only reference for the box design. Use the provided little knight reference as the only reference for the character, armor, drawing tube, sword, and hook rope. Generate a photoreal live-action cinematic image. This is a low-angle wide horizontal shot. The camera is placed very close to the ground, at grass height. The viewpoint feels close to the miniature knight’s scale, while still clearly reading as a giant real everyday environment. This is a wide environmental shot, not a close-up. The main background is the exterior wall of a giant green outdoor trash bin. The trash bin occupies most of the background and reads like a huge green wall. The bin has realistic outdoor plastic material. Show slight aging, tiny scratches, shallow dents, wear, and natural surface variation. The bin surface includes realistic vertical structure lines and slight unevenness. The ground is real outdoor grass. The foreground and lower half of the frame are dominated by tall grass. The grass must be photoreal, not illustrated and not painterly. The grass blades vary in length, direction, bending, thickness, and overlap. The layering is clear, with the nearest grass larger and softer, and the slightly farther grass more readable. The grass is very tall. The grass heavily occludes the little knight. The little knight is almost completely hidden below the neck by the grass. Only the helmet, the area above the neck, the upper shoulder line, and a very small amount of upper chest armor are clearly visible. Do not expose the full body. Do not clearly show the waist or legs. The overall read is that the grass feels like a huge field around him. The little knight is miniature in scale. He is a very small knight inside a giant everyday world. Keep realistic adult body proportions. Not a big-headed doll. Not chibi. The head size must feel natural. The body proportions must remain slender, believable, and realistic. The little knight stands on one side of the frame near the base of the trash bin. The box is placed on the opposite side of the frame. There is a clear spatial gap between the little knight and the box. Do not place the little knight directly beside the box. Do not make him touch the box. The image should clearly read as the little knight being in the grass at some distance, observing or moving toward the box and the environment. The little knight is standing still or in a very subtle forward-moving pose. His body faces toward the box. His helmet also faces toward the box. The action reads as quiet, focused, and cautious. Not an exaggerated action. Not a combat pose. More like he is studying the environment ahead. The little knight’s armor, helmet, drawing tube, sword, and hook rope must 100% match the provided character reference. Strongly emphasize the armor’s key feature: only one half of the armor has clearly visible pink floral engraved decoration. The other half is cleaner polished metal plate armor. This contrast between “one engraved half” and “one cleaner metal half” must read clearly. The engraved area must be visible. The metal has realistic highlights, slight wear, subtle seams, and convincing weight. The drawing tube on the back is clearly visible. The sword and hook rope are present, though they do not need to be fully exposed because of the grass occlusion. The box must 100% match the provided box reference. It is a normal corrugated cardboard box. The box is open at the top. It sits on the grass near the trash bin. Its size clearly reads as larger than the little knight. Keep the castle graphics visible on the box sides. The box surface has realistic cardboard texture, folds, slight wear, and natural sunlight response. The lighting is natural outdoor daylight. Most of the scene is in shadow. The area in front of the trash bin and much of the grass are primarily shadowed. Only a localized patch near one side of the frame is lit by sunlight. This bright patch falls across part of the trash bin, part of the grass, and the box. The lighting structure should clearly read as: broad shadow with one smaller sunlit area. The light is not evenly spread across the whole image. The lit area is soft but clearly defined. The shadows are realistic and layered. The little knight is mostly in the shadow area. The armor edges catch only a small amount of natural reflected light. Focus falls mainly in the mid-ground. Both the little knight and the box should be readable and clear enough. The nearest foreground grass can be softer and more out of focus. The trash bin remains readable in the background, but it should not overpower the subjects. Use shallow to moderately shallow depth of field. Keep the overall image grounded in realistic cinematic photography. Maintain a real-camera look. Add subtle film grain. Add slight sensor noise. Add mild lens softness. Add slight chromatic aberration. Add a very light white bloom. Keep the result natural and believable, not overly clean, and not like a toy advertisement.