AI limitations and human review

AI output is probabilistic. The same input can produce different results, and a visually strong page can still contain factual, textual, continuity, or rights-related errors.

Common limitations

• Character faces, clothing, props, and environments can drift between panels.
• Hands, small objects, crowded scenes, and complex poses can contain visual errors.
• Text drawn inside an image can be misspelled or unreadable.
• Panel continuity, reading order, and speaker attribution can require manual correction.
• A generated result can resemble familiar visual material even when that was not intended.
• Video timing, motion, audio, and lip synchronization can require additional editing.

Review before delivery

Check each panel at full size, proofread all text outside the image-generation step when possible, compare recurring characters against references, and verify export files after download. High-stakes educational, legal, medical, financial, or safety information requires qualified human review.

The Verified Comic Evidence section records observed results and limitations for specific controlled runs; it is not a guarantee that every generation will match those results.