hyperframes-media

Generate text-to-speech narration and transcription for video compositions by running the tts and transcribe commands, which produce voiceover audio from text and caption transcripts with timestamps using Whisper.

Remove backgrounds from video or images to create transparent overlays by running the remove-background command, which uses a u2net model to produce clean foreground cutouts suitable for composition layering.

You need to download models before transcribing speech or generating TTS audio, as each command downloads its required model on first run and caches it under the ~/.cache/hyperframes directory for subsequent use.

You can chain text-to-speech generation and transcription to automatically create captions by running the tts command to produce audio, then passing that audio into the transcribe command to generate timestamped caption data.

Configurable voice options and language settings are supported for voiceovers and caption transcripts, allowing you to specify TTS voice attributes and transcription language parameters through the tts and transcribe subcommands.

Limitations when using local models for background removal and speech transcription include the requirement of initial model downloads on first run, local storage usage within the ~/.cache/hyperframes directory, and dependency on the u2net and Whisper model capabilities.