MIND FORGE AI / PRACTICAL CREATION GUIDE
AI Video Prompt Library: Camera Motion, Product Shots & Common Fixes
Build a prompt around a subject action and a camera instruction. These original examples cover five motion patterns and include checks for common problems; they are suggestions, not measured H3 outputs.
Your practical workflow
- Decide whether the source image already defines appearance, lighting and composition.
- Choose one motion pattern and replace the subject with your own source material.
- Use a short draft, inspect the full clip and change one instruction at a time.
Text to video and image to video need different briefs
For text to video, define the subject, setting, light and movement. For image to video, the source already carries much of the visual brief, so focus on what changes over time. Avoid instructions that contradict the source, such as a stationary car surrounded by obvious motion blur.
The prompts below are original editorial starting points. Model behavior differs; terms from another model's documentation do not establish that H3 will reproduce the same motion.
Suggested prompts, not measured outputs. Select the text to copy it and adapt it to your own material.
Locked camera / product reflection
Gentle push-in / landscape
Lateral tracking / walking subject
Small arc / simple object
Environment only / original illustration
Diagnose the result before making the prompt longer
Review the whole clip, then identify whether the problem is in the source image, the requested action, the camera or fine detail. More words can add conflicting requirements. Keep a copy of the prompt and settings for each attempt.
On a narrow screen, scroll the table horizontally.
| Observed problem | First change to try | What to inspect |
|---|---|---|
| Subject cropped by a push-in | Use a locked camera and a wider source | Cap, feet, base and frame edges |
| Unwanted camera drift | Choose one explicit camera instruction | Background position and horizon |
| Object changes shape | Reduce rotation and use a clear source | Subject at start, middle and end |
| An unexpected cut | Simplify to one scene and one action | The full clip, not only a thumbnail |
A starting prompt to adapt
Editorial suggestion, not a verified generated result.
Review the result before another attempt
Inspect motion, framing, subject identity and any text. A prompt can guide behavior but cannot guarantee exact geometry or an artifact-free result.
Common questions
Should I write image-to-video prompts in English?
This library provides English and Chinese versions so you can try either. We have not measured which language works better across the site's models; compare your own task without changing other settings.
Will these prompts guarantee the requested movement?
No. They are suggested briefs, not validated presets for every model. Start with a short attempt and assess the actual output before extending duration or spending more credits.
References
External documentation supports the specified platform rules, plan details or terminology, not performance claims about this site's models.