Get the best results from the AI video and image generator with these simple practical tips
1. Keep movements simple & subtle
Avoid overly complex or exaggerated actions like backflips, intricate hand gestures, or full-body acrobatics.
For animals: opt for basic actions like walking, sitting, or looking around.
For humans: simple gestures like turning the head, light hand movements, or shifting weight between feet tend to render better than detailed choreography.
2. Frame the subject wisely
Mid-shots (waist-up) or bust shots (chest-up) for humans can help minimize wonky leg and foot movement.
For animals, framing them in natural settings like lying down or sitting reduces the risk of unnatural movement.
3. Use poses where the body isn’t blocked or covered.
Try to avoid poses where hands cover faces, arms cross bodies, or limbs overlap awkwardly. Keeping body parts separated in space helps the AI "read" anatomy better.
4. Prioritize shorter clips
The longer the video, the more likely there’ll be glitches or inconsistencies, like jitter or unnatural limb movements.
Generate shorter clips and loop or stitch them together later in post-production.
5. Stylize when possible
Lean into non-photorealistic or stylized outputs (e.g., animation, painterly looks, or cinematic lighting) to mask imperfections.
A cartoonish or surreal visual style can help "justify" slight weirdness.
6. Use prompt engineering
Be super specific in your prompts. Instead of "dog running," try "a golden retriever walking calmly in a grassy field, viewed from the side, natural movement."
Use specific prompt details:
- Include clear subject, medium, environment, lighting, color, and composition details in your prompts
- Use specific numbers instead of plural words (e.g. "three cats" instead of "cats")
- Focus on describing what you want rather than what you don't want
Updated