NOVABERDocs

Loading search index…

    Documentation menu

    App reference

    Cutout and watermark removal reference

    26.11.1006

    Novaber · Content reviewed

    On this page

    Batch cutouts and GPU watermark removal on Windows for images, short MP4 clips, and ordered images. Run either operation or both, then preview and export transparent PNGs, masks, or restored media.

    Capabilities and input / output

    • Batch background removal with BiRefNet, SAM3 text prompts, or SAM2 point selection

    • GPU-accelerated SLBR removes visible watermarks from images, short clips, and ordered images with automatic detection or manual selection

    • SAM2 subject tracking in short clips and ordered images, with keyframe corrections and frame-by-frame preview

    • Compare with the original and export transparent PNGs, masks, restored images, or MP4, depending on the operation

    Workflow

    1. Import media

    Import an image batch, one MP4, or ordered images. Set the clip range and extraction frame rate for video. Clips and ordered images support up to 300 frames.

    2. Choose an operation

    Choose cutout, watermark removal, or both. SLBR runs on the GPU and downloads weights on first use. Combined processing removes watermarks before cutout for image batches, video, and ordered images.

    3. Track and refine

    Use SAM2 to mark a subject, refine keyframes, and preview short clips or ordered images. Combined processing removes watermarks first, then tracks the subject in the restored frames to create masks and transparent results.

    4. Compare and export

    Compare the original and results, then export a ZIP for the selected operation. Watermark removal on video also includes a restored MP4.

    Requirements

    • GPU: NVIDIA GPU required
    • VRAM: 8 GB+ VRAM recommended
    • Images and models: Image downloads ~16 GB

    Review your first result

    These checks are suggestions based on published inputs and outputs. Use a small sample to verify processing, preview and export together; no measured accuracy or speed is provided here.

    Identify batch images, short MP4 or ordered images. Inspect edges and transparency for cutouts and repaired areas for watermark removal. For combined tasks, inspect repair before cutout. Unzip the export and distinguish masks from transparent PNGs. Keyframe correction for videos and sequences is not a batch-image operation.

    Version, limitations and license

    Current catalog version: 26.11.1006.

    Third-party names and trademarks belong to their owners. The SLBR authors provide public access to the code and weights. The code license and rights to redistribute the weights remain unconfirmed.

    Product details and installation

    Batch cutout interface guide

    Choose by media, model and intended output

    GoalSelection and operationConfirm before running
    Automatic batch background removalBiRefNet; inspect representative subjects firstTransparent images or masks for delivery
    Select an image subject with textSAM3; describe the intended objectPrompt matches the actual target
    Select foreground using pointsSAM2; inspect keep/exclude point predictionsPredicted region matches the original
    Track subjects in a short MP4 or ordered imagesSAM2; review keyframes and individual framesOrder, clip, extraction settings and frame count
    Remove visible watermarksSLBR automatic detection or manual selectionLocal model readiness and repair region
    Repair watermarks and retain a transparent subjectCombined processing, repair before segmentationReview repaired and segmented outputs separately

    This page matches the catalog version above. Clips and ordered images support up to 300 frames. Ordinary image batches do not use the video keyframe workflow. Check matching documentation for controls introduced by another version.

    From first processing to output review

    1. Establish the model and goal with a small sample

    Open the app from Novaber, confirm service and model readiness, import authorized material and keep originals. Choose a few images with representative thin edges, holes or complex backgrounds; define clip and frame limits for video. Select a route from the table before processing.

    2. Inspect each processing layer

    Compare cutouts with originals for missing subjects, retained background and edge defects. For SAM2 temporal tasks, inspect frames for drift and correct keyframes. Inspect watermark repair regions and texture. Review repair and segmentation independently for combined tasks.

    3. Check exports by delivery purpose

    OutputPurposeCheck after reopening
    Transparent PNG / RGBACompositing and layoutAlpha, edges, holes and complete subject
    MaskInspect or reuse subject regionsCorrespondence with the input
    Restored imageReview watermark repairResidue, texture and subject detail
    Restored MP4Video watermark resultsClip, continuity and repair region

    Export the ZIP for the selected operation and inspect extracted count, formats and content. Restored MP4 applies to video watermark removal, not every task. Follow the batch guide for steps and app FAQs for residue, transparency and download issues.

    Prepare the environment

    For NVIDIA GPU workflows, check driver major >= 575 and Windows driver >= 576.57 for the project’s CUDA 12.9.1 baseline. See the driver and CUDA baseline for version comparison, unknown results and container checks. Check the selected app’s GPU model, VRAM and model requirements separately.

    Use the versioned setup skill for assisted checks.

    Sources