Community Help

Speech models can infer generic phrases from ambiguous input such as silence, music, applause, noise, or television transitions. VoxRelay's current build filters common false phrases, probable non-speech, low-confidence segments, and repeated overlap, but no model can eliminate every incorrect caption.

Use the latest VoxRelay version, select the cleanest playback source, keep speech at a clear level, and avoid unnecessary audio processing or background music when practical. Include the app version, audio source, and a description of the surrounding sound when reporting a repeatable false-caption problem. Do not post private recordings or transcripts.