Blurbly video test · 2026-10-21
AI UGC candle video: corrected single-frame generation
A fictional AI presenter gestures beside an amber candle in one full-frame scene. This five-second correction has no audio stream or split screen; unwanted mouth movement remains. Internal AI demonstration, not a customer review or approved ad.
Internal generation test with a fictional product and fictional AI presenter. No real customer testimonial, measured ad performance or platform approval. Initial entry used the default Review style despite the no-style request; the correction supplied no style reference. This button loads the correction brief, not the rejected first brief.
Make one with your own product
Add a clear photo of your product and describe the clip. You see the plan and the price before anything is generated.
The source and actual output
We started from our own AI-generated fictional candle photo in Blurbly AI, generated a fictional presenter and storyboard, then reviewed an actual UGC-style video. The first take failed our framing check; a single-frame retry removed the split screen. Both attempts and their full costs are recorded. No customer-uploaded image was used.
- Output
- Corrected raw clip: 720 × 1296, 5.00 seconds, silent MP4. The initial automatic cut was 1080 × 1920.
- Model
- Wan 3.0 Reference for the first take; Wan 3.0 first-frame generation for the correction. Reference images: Nano Banana Pro.
- Actual test cost
- 1,560 credits across both attempts: 900 for three reference images, 330 for the rejected video, and 330 for the correction.
For this case, the button loads the generated presenter start frame shown below, including the candle and fictional AI presenter. The original product-only photo stays above for comparison. The new plan can differ; this is a starting point rather than an exact replay.
What we reviewed
- The first 330-credit video introduced a split screen around the third second and an animated speaking mouth. Its automatic cut added captions and music despite the original brief. This rejected attempt remains in the test record.
- For the correction we used one generated storyboard frame as a first-frame input, removed the style reference and storyboard grid, and turned Sound and Prompt enhance off. No external retouching or video editing was applied to the corrected clip.
- Five sampled frames from the corrected clip show one continuous full-frame scene, the same fictional presenter and one candle. The downloaded file has no audio stream. The blank label and amber jar remain recognizable; fine geometry is not certified.
- The presenter still moves her mouth and hands. The request for a completely closed mouth was not fully followed, even though the file is silent. This is a reviewed AI demonstration, not a fully approved ad or a real customer review.
Exact input brief
Animate this single start-frame photo as exactly ONE continuous 5-second shot. The adult AI presenter is fictional. Keep the same woman, plain cream shirt, table, single amber candle and bedroom, with the composition fixed throughout. Her mouth stays completely closed; she gives a tiny closed-mouth smile and one very small relaxed open-hand gesture safely beside the candle, never touching it or the flame. Keep the whole candle and both hands visible. Static medium framing; only slight natural body movement and flame flicker. No camera cuts, no changes of angle, no zoom, no new scene, no inset, no split screen, no collage, no extra product, no captions, no on-screen text, no added logo, no voice, no music and no sound. Keep the blank kraft label blank and preserve the jar geometry. This is a silent AI-presenter demonstration, not a real customer review or a product performance claim.
This records the request, not a promise that the model followed every instruction. Review the output above before adapting it.
Keep exploring
All video tests · Open the matching studio · Step-by-step tutorials