Image/Video generations not accurate

Image and video generation relies on AI models that interpret prompts differently. Results can vary depending on:

Even small prompt changes can lead to noticeably different results.

What to try first

For the best results, we strongly recommend:

These models:

Avoid switching models mid-project, as this often causes inconsistent faces or styles.

For the best results, we strongly recommend:

These models provide:

⚠️ Video generation uses more credits than images.

For better video results:

Keep prompts concise but specific
Avoid very long descriptions
Clearly specify:
- Scene
- Style
- Camera motion (if needed)
- Expected duration (if available)

Overly complex prompts often reduce accuracy.

Vague prompts often lead to unpredictable results.

Instead of:

Try:

“Keep the same face, same ethnicity, same facial features, no changes to identity.”

The more specific you are, the better the outcome.

For images with people:

If you need precise results:

This reduces wasted credits and improves consistency.

Image and video generations consume credits once processed, even if:

Credits cannot be refunded or re-added. This behavior is expected.

The following are normal technical limitations:

These are not considered bugs.

If generations consistently fail or don’t start at all, contact support and include:

This helps us identify genuine issues faster.