ffmpeg-usage

Use the command ffmpeg -i input.mov -c:v libx264 -c:a aac -strict experimental output.mp4 to re-encode with H.264 video and AAC audio. For a universal conversion from any format, use libx264 with CRF 23 and AAC at 128k bitrate.

Generate a color palette first with palettegen, then apply it with paletteuse in a second ffmpeg pass for accurate colors. Limit the frame rate to 10 fps and scale width to around 480 pixels to keep the GIF file size manageable.

For MP4, MOV, or MKV files, use the concat demuxer with a file list and -c copy to avoid re-encoding. The concat protocol only works with MPEG-TS, MP3, and AAC formats, while the concat filter is needed when inputs have different codecs or resolutions.

Yes, transcription runs through a local HTTP API at 127.0.0.1:18923 using the SenseVoice Small model, so no internet connection is required. The model must be downloaded first (about 239 MB), and ffmpeg is needed for non-WAV audio formats.

CRF 23 is the default balance of quality and size, 18 is visually lossless, and 28 produces smaller files with acceptable quality. The scale runs from 0 (lossless) to 51 (worst quality), and higher presets like veryslow improve compression efficiency.

Issues occur when concatenating files with mismatched codecs, resolutions, or timebases using stream copy. Re-encode with the concat filter instead, which normalizes all inputs into a single consistent output stream.