Audio Formats

Required Format

Parameter	Value
Encoding	PCM (signed 16-bit little-endian)
Sample rate	16 000 Hz
Channels	Mono (1 channel)

The server buffer uses 32 KB slots — at 16 kHz mono 16-bit PCM, 1 second of audio = 32 KB.

Converting with FFmpeg

ffmpeg -y -i input_file \
  -ac 1 -acodec pcm_s16le -ar 16000 \
  -f wav output_file.wav

Use - for stdin/stdout to pipe directly without writing a file:

ffmpeg -i input.mp4 -ac 1 -acodec pcm_s16le -ar 16000 -f wav - | your_streaming_script

Chunk Sizes

Chunk size	Duration at 16 kHz	Notes
4 KB	~125 ms	Low latency
32 KB	~1 s	Matches server buffer slot size
64 KB	~2 s	Larger chunks, higher latency

Smaller chunks produce faster responses. Larger chunks may improve accuracy at the cost of latency.