Use this guide to diagnose what went wrong with a generated video and apply the right fix. Diagnose before re-rolling — re-submitting the identical video prompt usually repeats the failure.
A note on negative prompts, referenced in the fixes below: a negative prompt is a separate input field, supported by some video models, where you list things you do not want to appear (for example "blurry, extra fingers"). If your video model has no negative prompt field, phrase avoidances positively in the main prompt instead: rather than "no blur," write "sharp focus throughout."
The list below pairs each common video generation problem with its quick fix:
Problem: Static or lifeless output
Fix: Add an explicit camera move and action verbs to the video prompt.
Problem: Wrong anatomy (extra fingers, warped limbs, broken faces)
Fix: See "How do I fix wrong anatomy" — reduce the number of people, describe poses clearly, use anatomy-targeted negative prompts.
Problem: Distortion, morphing, or warping
Fix: See "How do I fix distortion" — simplify the action, shorten the clip, avoid emotionally loaded or ambiguous wording.
Problem: Video ignores my reference image / person looks different
Fix: See "Why doesn't my video match my reference image" — strengthen the reference image and describe motion only, not appearance.
Problem: A real person's face won't animate properly or loses likeness
Fix: See the image-to-video tips — use a sharp, well-lit, close-up source photo with nothing covering the face, upscale it first if soft, keep the motion subtle, and never describe the face in the video prompt.
Problem: Output looks generic or stock-footage-like
Fix: See "Why does my video look generic" — replace vague adjectives with specific, filmable details.
Problem: Prompt details ignored
Fix: See "Why is the model ignoring my prompt" — move key details earlier in the video prompt, cut competing descriptors.
Problem: Blurry or soft output
Fix: See "How do I fix blurry video" — request "sharp focus" positively and add "blurry, out of focus, low resolution" to the negative prompt if available.
Problem: Wrong speed/pacing
Fix: State it directly in the video prompt: "slowly," "in slow motion," "rapidly."
Problem: Generation blocked by content filter
Fix: Rephrase the flagged element — do not re-submit the identical video prompt. Common triggers: naming celebrities or real people (describe the action instead and let the image carry the identity), violent or explicit wording, and words like "famous" or "lookalike." Note: NSFW (not-safe-for-work, meaning sexual or explicit) content of real people from uploaded photos is blocked by policy and cannot be rephrased around.
Problem: NSFW output blocked or degraded quality on explicit content
Fix: Two steps that significantly reduce blocks and improve output quality for NSFW generations:
1. Age anchoring — explicitly state the subject's adult age in the prompt (for example, "a 25-year-old woman"). This reduces ambiguity that can trigger age-related safety filters.
2. Clean reference images/videos — ensure any uploaded reference image or video does not contain underage individuals or content that could be flagged as such. Even if your prompt is compliant, a flagged reference can cause the generation to fail or be blocked.
Negative prompt tip: name the exact flaw you saw — "extra fingers, warped hands, flickering, morphing" — rather than pasting a giant generic block, which can fight your positive prompt and miss the actual problem.