Based on the reference image provided by the user, after in-depth analysis from the perspective of a cinematic storyboard artist, expand into 3 storyboard prompts with different narrative functions.
Core concept: This is an img2img workflow. Each prompt must explicitly state at the beginning "based on the character/scene in the reference image," so that the AI generation tool knows to transform the shot based on the original image, rather than generating from scratch. The prompt itself must include a reference instruction to the original image.
Workflow
Step 1: In-depth image analysis (anchor shot extraction)
Carefully observe the image provided by the user, extract visual information according to the following dimensions, and form an "anchor shot" parameter table. All subsequent storyboards must remain consistent with these parameters.
| Dimension | Extracted content |
|---|---|
| Aspect ratio | The original image's aspect ratio (such as 16:9, 4:3, 1:1, 9:16, 3:4, etc.); subsequent generations must remain consistent |
| Character appearance | Gender, perceived age, face shape, body type, hairstyle/hair color |
| Clothing | Material, color, style, accessories, degree of wear |
| Scene/worldview | Era, type of location, environmental details (architecture, vegetation, weather, etc.) |
| Visual style | Realistic/animation/cyberpunk/oil painting, etc.; rendering texture (film/digital/3D, etc.) |
| Color tone | Warm/cool tendency, saturation, dominant color, color grading name (such as teal and orange) |
| Lighting | Light source direction, color temperature, contrast between light and dark, lighting type (Rembrandt lighting/butterfly lighting/natural light, etc.) |
| Composition | Current shot size, character position, foreground/background relationship |
| Props | Material, color, and state of key objects |
| Emotional atmosphere | The emotional tone conveyed by the image (loneliness/warmth/tension/serenity, etc.) |
Output the analysis results to the user in a concise table format as an anchor shot record.
Step 2: Determine the narrative function
Read references/narrative-function-guide.md, and analyze the role of the original image in the narrative according to the judgment decision tree:
- Establishing
- Introduction
- Climax
- Turning Point
- Daily Moment
- Confrontation
Based on the judgment result, determine the specific direction for expanding the 3 storyboards. Briefly inform the user of the judgment process and conclusion.
Step 3: Generate 3 storyboard prompts
Read references/shot-language-reference.md, and generate 3 prompt segments according to the following fixed role division:
Storyboard role division:
- Storyboard 1 — Spatial exposition: Long shot/extreme long shot, revealing the complete space and worldview
- Storyboard 2 — Emotional close-up: Medium close-up/close-up, capturing the character's inner state
- Storyboard 3 — Narrative progression: Free shot size, showing events before and after in time
Prompt writing specifications:
- Word count: 80-120 characters per paragraph (Chinese characters)
- Reference the original image: The beginning of each prompt paragraph must include a reference instruction such as "Based on the reference image," or "Maintain the appearance features of the person in the reference image," clearly telling the AI generation tool that this is a variation based on the original image, not generated from scratch
- Shot opening: After the reference image instruction, immediately follow with the camera movement/shot scale (selected from the sentence pattern library or created yourself, such as "slowly pushing in medium close-up" or "extreme high-angle wide shot")
- Content organized by formula：
[引用原图指令], [镜头类型&角度], [人物完整外观+动作], [有层次的场景描述], [光线设计], [色调/色彩分级], [情绪氛围] - Character description: Each paragraph must fully write out the character's appearance and clothing, and must not omit them or write "same as before" or "as above"
- Consistency: Color tone, light direction, visual style, and time period must remain strictly consistent with the anchor shot
- Variation: The shot scale, angle, and composition of the 3 storyboard shots must be clearly different
- Specificity: Use specific visual elements instead of vague adjectives (do not write "beautiful," write "dyed amber by the sunset")
- Aspect ratio: All storyboard shots must maintain the same aspect ratio as the original image
Step 4: Quality self-check
Check each item before output; if it fails, revise and then output:
- Each paragraph is 80-120 characters
- Starts with camera movement/shot scale
- Character description is complete and consistent across frames
- Visual style, color tone, and lighting are consistent with the original image
- The 3 storyboard shots have clearly different shot scales (wide shot + close-up + free shot scale)
- No omissions such as "same as before," "as above," or "same as the first image"
- Use specific visual elements, with no vague adjectives such as "beautiful," "pretty," or "good-looking"
- Each paragraph can be used independently and copied and pasted directly into an AI generation tool
- The aspect ratio is consistent with the original image
Output format
Directly output 3 paragraphs of plain text prompts, with one blank line between paragraphs. Do not add tags, numbering, parameter symbols, or markdown formatting. Each prompt paragraph itself contains the reference image instruction, so the user can use it directly with the original image upon receiving it.
After outputting the prompts, attach brief parameter suggestions:
- Aspect ratio: Consistent with the original image (such as 16:9, 3:4, etc.)
- Reference strength/similarity: Recommended 0.6-0.75
Example output:
Based on the reference image, maintaining the same character appearance and visual style. An aerial wide shot slowly pans sideways, an industrial port shrouded in mist reveals its outline in the morning light, rusty orange-red gantry cranes are lined up along the edge of the gray-blue dock, the thin middle-aged man in the reference image wearing a dark blue work jacket and with a crew cut stands alone at the end of the pier, his figure appearing small among the huge container stacks, the sea surface reflects the cold gray skylight, white smoke from a distant chimney is blown by the wind toward the left side of the frame, the overall color tone leans cold cyan-gray with a light orange touch, and a desolate industrial feeling of early morning fills the scene
Based on the reference image, keeping the same person and color grading. A shallow depth-of-field close-up of the face slowly pushes in; the gaunt middle-aged man in the reference image, wearing a dark blue work jacket and with a buzz cut, lowers his head slightly, with faint bluish-black shadows beneath his deeply sunken eye sockets, a rough-fingered hand holding an unlit cigarette, the corners of his mouth drawn tight into a line, the left side of his face lit by the cold gray morning light reflected off the sea, the right side sunk into blue-black shadow, the background completely blurred into vague spots of light from the harbor, and within the cold cyan-gray tones a suppressed exhaustion and silent tenacity show through.
Based on the reference image, same character and cinematic tone. A side-tracking medium-wide shot moves slowly forward along the pier; the gaunt middle-aged man in the reference image, wearing a dark blue work jacket and with a buzz cut, walks away from the camera toward the depths of the dock, carrying a worn canvas tool bag in his left hand, a few patches of rust-colored dust on the back of his work jacket, the wet concrete pavement beneath his feet reflecting the skylight, stacked shipping containers on both sides forming a narrow passage, a yellow forklift parked in the fog in the distance, cold cyan-gray tones enveloping the entire scene, as a new day's labor in the early morning is about to begin.