To reduce face distortion in a Burning Bridges AI video, start with a sharp, near-front portrait showing one person's head and shoulders in even light. Check whether the face stays clear in the source preview before changing output resolution: resolution affects export detail, while the photo supplies the likeness for a fixed 15-second performance.
| Symptom | First check | Practical adjustment |
|---|---|---|
| Features shift during movement | Face angle and visibility | Use a more direct view with eyes, nose, and mouth unobstructed |
| Face looks soft throughout | Focus and image compression | Choose a sharper original instead of an enlarged crop |
| One side looks unlike the source | Uneven light or strong shadow | Prefer balanced light across the face |
| Output is blocky but the likeness is stable | Resolution tier | Consider a higher export tier only if more detail is needed |
Choose a portrait with visible facial detail
Use a single person facing close to the camera, with the full face and head clearly visible. Strong profiles, an extreme head tilt, sunglasses, hands over the face, or hair covering key features leave less visible information. A head-and-shoulders frame is a practical starting point; a tiny face in a distant or crowded photo is harder to inspect and may not provide the same clarity.
Prefer a focused image over a screenshot or a small face enlarged from a larger photo. Avoid heavy filters that reshape facial proportions. If you have several options, compare them at normal viewing size and select the one where the eyes and facial outline are clearest.
Change one factor at a time
If the result is completed but the face drifts, keep the same output tier and try a better source portrait on a later attempt. Changing angle, lighting, crop, and tier all at once makes it difficult to identify which change mattered. Review the whole clip, not only its opening frame, because movement can reveal inconsistencies that are not obvious in a still preview.
The template itself is not prompt-directed, so adding imagined instructions cannot correct a photo through a user-facing free-text prompt. The prepared scene, performance, and audio remain fixed; the source portrait is the main relevant variable documented for this workflow.
When changing resolution helps
The configured clip is 15 seconds in 16:9. The available tiers are 480P for 50 credits, 720P for 100, and 1080P for 200. A higher tier may be relevant when the output lacks export detail, but it does not change the face angle, lighting, or identity source. Check current options at the pricing page before spending more credits.
If a video completes but the likeness is unstable, use the portrait guide and improve the source. If no video is produced or a job is marked failed, check the job and credit status and contact support; that is a render issue, not face distortion.