What this guide helps you do
Write better AI image-to-video prompts for documentary scenes
Key takeaways
- Begin with the shot's narrative job.
- Lock appearance in a still before directing motion.
- Use references and continuity receipts instead of repeating vague adjectives.
Film evidence
Prompt for a shot's job, appearance, and motion separately
A documentary prompt should inherit an approved beat and evidence boundary. For this explainer, one shot might orient the Himalayan range at geological scale, another isolate wind across Everest's summit, and another evoke the human expedition layer without pretending a generated vintage image is archive footage.
Describe what the image looks like separately from what moves. Include subject, setting, composition, lens behavior, light, duration, continuity state, prohibited details, and the intended transition. Then review the generated result against the sequence, not the prompt alone. A syntactically compliant shot can still be misleading, repetitive, or wrong for the edit.
Rising Giants · 00:50 · Explainer cut. This is a finished first-party Onira production, not customer proof or archive footage.
View the full film and production notesWhat to watch for
- One explicit editorial purpose per shot
- Motion that supports rather than changes the factual claim
- Visual style never obscures reconstruction status
Section 1
Define the editorial job before the visual
Write one sentence explaining what the viewer must understand or feel during the shot. The same line of narration might need orientation, evidence, scale, contrast, human detail, or a transition. That purpose determines the subject and framing more reliably than a style adjective.
Avoid asking one shot to establish a city, introduce three people, demonstrate a process, and deliver emotional climax. Divide complex beats into a sequence of readable images.
- +Narration segment and exact duration.
- +One visual purpose.
- +One focal subject or relationship.
Section 2
Write the image prompt
Describe the subject, setting, period, materials, wardrobe, spatial composition, camera position, lens character, depth, light, color, weather, and documentary tone. Add negative constraints only when they prevent likely failure.
Use concrete visual nouns instead of stacked praise such as stunning, epic, cinematic, ultra-detailed. The model needs a scene to construct, not a review of the hoped-for result.
- +Who or what is visible and where.
- +Period, geography, materials, and environmental conditions.
- +Framing, lens, light, texture, and restrained style.
Section 3
Write the motion prompt
Assume the approved still already defines appearance. Describe the main subject action, camera behavior, environmental motion, speed, and end state. Keep the physical request simple enough to remain coherent for the shot duration.
Good documentary motion often comes from small changes: cloth moves in wind, a hand completes one craft action, smoke drifts, light changes, or the camera reveals scale. Motion should not invent a new scene halfway through the clip.
- +One principal action.
- +One camera move or a locked camera.
- +Subtle secondary motion and a stable ending.
Section 4
Carry continuity across shots
Store accepted character, subject, wardrobe, setting, lighting, and style references. Each later prompt should use the relevant reference rather than attempting to recreate identity from prose alone. Track whether a shot must begin from the preceding final frame or can cut cleanly to a new setup.
Review continuity at sequence speed. Individual frames can look attractive while identity, screen direction, time of day, or geography drifts across the edit.
- +Accepted reference images and descriptive identifiers.
- +Continuity requirements by adjacent shot.
- +Sequence-level review after generation.
Working standard
Publication checklist
- 01Every shot has one editorial purpose and target duration.
- 02The image prompt defines appearance with concrete detail.
- 03The motion prompt does not redundantly redesign the frame.
- 04Actions and camera direction are physically simple.
- 05Accepted references are reused for recurring elements.
- 06Continuity is reviewed in the assembled sequence.
Primary references
Sources and further reading
Policy and model capabilities change. These sources were reviewed on July 11, 2026; open the current official page before making a production or publication decision.
Related production guides
Questions
What should an AI documentary video prompt include?
Include the shot's editorial purpose, subject, setting, period, composition, camera, light, and constraints. For motion, describe one action, camera behavior, environmental movement, and end state.
Should image and video prompts be separate?
Yes. The image prompt defines what the shot looks like. After approving that anchor, the video prompt should primarily define what moves and how the camera behaves.
How do I keep a character consistent?
Use accepted reference images and a stable character description, keep wardrobe and lighting receipts, and review the complete sequence. No current workflow should promise perfect identity in every generation.