Prompting Guide
I2VA uses the uploaded image as the exact first frame at 0.00s. Do not waste the prompt re-describing the image. Focus on what happens next.
first frame → action → camera motion → reaction/result
Required Prompt Format
Keep these field names unchanged.
integrated_multimodal_description: [Shot 1] ...
overall_soundscape: ...
non_diegetic_music: ...
Writing the Main Prompt
Inside `integrated_multimodal_description`, describe events in chronological order:
[Shot 1] {style + framing}. Preserve {important first-frame details}. {subject begins action}. The camera {camera motion}. {action continues + secondary motion}. {final pose/reaction/result}.
Example
integrated_multimodal_description: [Shot 1] Cinematic live-action, medium close-up. Preserve the woman's face, hairstyle, clothing, seat position, and train interior from <Picture 1>. She slowly lifts her eyes from the letter and turns toward the rainy window. The camera pushes in slowly. Her reflection slides across the glass as she folds the letter and lowers it into her lap, ending with her gaze fixed outside.
Rules
- Treat `<Picture 1>` as the locked first frame.
- Describe motion, not the image itself.
- Preserve important identity, clothing, props, colors, layout, and lighting.
- Write actions as visible physical steps instead of vague commands.
- Keep movement continuous and physically plausible.
- Use one clear camera move when possible.
- For short clips, one coherent shot is usually easier to control.
- End with a clear pose, reaction, object state, or composition.
Better Motion Description
Weak: She stands dramatically.
Better: She leans forward, plants both feet, pushes against the table with one hand, rises, and straightens her shoulders.
Camera Terms
`Push In` / `Pull Out` — camera moves forward/backward
`Zoom In` / `Zoom Out` — lens zoom only
`Pan Left` / `Pan Right` — camera rotates horizontally
`Truck Left` / `Truck Right` — camera moves sideways
`Tilt Up` / `Tilt Down` — camera rotates vertically
`Pedestal Up` / `Pedestal Down` — camera moves vertically
`Arc Shot` — camera moves around the subject
`Tracking Shot` — camera follows the subject
`Static Shot` — no camera movement
`POV` — point-of-view shot
`Roll Clockwise` / `Roll Counterclockwise`
`Shake Slightly` / `Shake Strongly`
Optional modifiers:
with small amplitude
with large amplitude
at slow speed
at fast speed
Example: The camera trucks right slowly as she walks toward the doorway.
Multiple Shots
`[Shot 1]` starts at 0.00s and does not need a timestamp.
Later shots require increasing timestamps:
[Shot 2] At 00:04.500, the camera cuts to a close-up of his hand gripping the railing.
Useful transitions:
the camera cuts to
the shot cuts to
the shot transitions to
the shot changes to
Only add cuts when they reveal a genuinely new angle, reaction, detail, location, or state.
Dialogue
Assign stable speaker IDs such as `(S1)` and `(S2)`.
The woman with a quiet voice (S1) says: <d>[English] I knew you'd come back.</d>
Inside `<d>...</d>`, include only:
[Language] exact spoken dialogue
Put speaker identity, emotion, and delivery outside the dialogue tag.
For voiceover:
The man (S1) says in an off-screen voiceover: <d>[English] I remember this place.</d> while his lips remain completely closed.
On-Screen Text
Put exact visible text in double quotes:
The sign reads "OPEN 24 HOURS".
Sound
`overall_soundscape`
Briefly describe environmental and physical sounds. Do not repeat dialogue here.
overall_soundscape: Soft rain hits the train window over a low rail hum. Paper rustles as the letter folds.
`non_diegetic_music`
Describe background score heard by the viewer but not the characters.
non_diegetic_music: Sparse piano with sustained low strings at a slow tempo, fading near the end.
If no background score is wanted:
non_diegetic_music: N/A
Music coming from a radio, phone, TV, musician, etc. is diegetic and belongs in the main scene description.
Best practice: concise, chronological, visual, and specific beats long descriptive prose.