The way I was converting my WAV from stereo to mono 16-bit and 16 kHz, ffmpeg was appending LIST-INFO metadata, and the DeepSpeech client is very particular about the WAV files it processes. I ran the same DeepSpeech native client command, and lo-and-behold, I actually started getting the output from the large file. using DeepSpeechClient ; using System ; using System.IO ; using System.Threading.Tasks ; namespace ds_dotnet { class Program { static async Task Main ( string [] args ) { DeepSpeech deepSpeechClient = new DeepSpeech ( "deepspeech-0.9.3-models.pbmm" ); var waveBuffer = new NAudio .