NOVABERDocs

Loading search index…

    Documentation menu

    Support and troubleshooting

    Cutout & Watermark Removal FAQs

    26.11.1006

    Novaber · Content reviewed

    On this page

    This page uses published catalog version 26.11.1006. Normal operations are covered by the app reference and batch cutout guide. Use matching documentation for other versions; candidate controls do not establish availability in older releases.

    Why do images, video and ordered images need different operations?

    Image batches process photos; short MP4 clips and ordered images also involve frame order, keyframes and subject tracking. Identify the media type before choosing an operation. Temporal keyframe corrections are not ordinary batch-image edits.

    The published workflow for this version limits clips and ordered images to 300 frames. Record the clip, extraction settings and actual error when investigating range or frame-count issues. Follow media and model selection.

    Why does a listed model still need preparation or a download?

    Supported does not mean locally installed or ready. Check app service and model state, separating incomplete images, pending weights and computation failures. SLBR downloads weights on first use; preparation time does not measure processing speed.

    After host-driver and container GPU checks, run a small task with the selected model. See environment questions and include the model and failing stage in feedback.

    What should I check for remaining watermarks or damaged subject detail?

    Compare with the original to locate the defect in the watermark area, background texture or subject. Automatic detection and manual selection have different scopes. Check coverage while avoiding unrelated regions, then review a small sample.

    Combined processing removes watermarks before cutout. Inspect repaired images and transparent results separately; the final cutout alone cannot verify repair. Keep originals and examples of defects. Complete removal is not guaranteed. See processing and output checks.

    Does a white or black PNG background mean transparency was lost?

    A viewer may display transparency over a solid color. Confirm the file is a transparent PNG rather than a mask, thumbnail or watermark-only output. Open it in software supporting alpha and place it over light and dark backgrounds.

    Inspect edges, holes and intended transparent regions in the actual export. If the background remains opaque, record input, model, operation and file type, then follow export checks.

    How do masks, transparent PNGs and restored images differ?

    A mask records subject regions. Transparent PNG / RGBA retains the subject and alpha. Restored images are used to inspect watermark repair. Choose the output by delivery purpose.

    Unzip exports and compare the selected operation, file count and contents. The published video watermark workflow also includes a restored MP4; that does not mean every task exports identical files. See the output reference.

    How do I report a completed task with unexpected download contents?

    Confirm task, media, model, operation and export options, then inspect the extracted files. Use a small reproduction and state whether the expected output is a transparent subject, mask, repaired image or video, and what was actually produced.

    Include desktop/app versions, completion state, exact error and relevant screenshots. Private originals are not required. Follow feedback guidance for privacy and submission.

    Sources