Disclaimer

1. Machine Learning & AI Accuracy

Tools utilizing machine learning models (such as the neural background segmentation model) are AI-assisted; results vary depending on subject complexity, edge contrast, lighting conditions, and image resolution.

Neural networks estimate alpha probability masks and may exhibit artifacts around fine details such as wispy hair, transparent glassware, or low-contrast backgrounds. Manual refine brushes, threshold adjustment, and edge feathering controls are provided to correct these variations.

2. Browser Memory & Hardware Limitations

Because all file decoding, buffer manipulation, and rendering execute entirely inside your device's web browser process, available system RAM and graphics memory (VRAM) directly constrain file processing capacity:

• Video Processing: Demuxing large video files (up to 1 GB) requires sufficient RAM for container parsing. On lower-memory mobile devices, processing very large video files may cause the browser tab to exhaust memory.
• High-Resolution Images: Canvas operations for images exceeding 50 megapixels may exceed browser canvas allocation limits.
• Vector Tracing: Tracing intricate photographic details into thousands of vector paths can generate large SVG files that require significant browser memory to render.

3. Audio Stream Copy vs Re-Encoding

Our video-to-audio extraction tool utilizes stream copy when saving AAC audio into M4A containers, meaning audio frames are extracted without re-encoding and with zero generational loss. However, converting audio to MP3 or WAV formats involves decoding and re-encoding or uncompressed PCM serialization.

4. Third-Party Codec and Format Compatibility

Certain proprietary formats or non-standard container codecs (such as HEVC/H.265 video tracks or Dolby AC3 audio) may not be natively decodable across all browser engines due to operating system licensing constraints. In such cases, the tool will alert you that the specific track cannot be read.