RecursiveAI.OneSpeech.Cli 1.1.0

Speech recognition for .NET

OneSpeech transcribes audio, times words, labels speakers and runs live sessions in managed .NET 10. Open-weight models (Whisper, Qwen3-ASR, FastConformer, Moonshine, Silero VAD, CAM++) run on OneModel; OneAudio decodes and resamples audio. No native inference runtime is required, and the packages are NativeAOT and trim clean.

OneSpeech is unaffiliated with other products or companies named OneSpeech.

using OneSpeech;

Transcript transcript = Speech.Transcribe("meeting.m4a");
foreach (Turn turn in transcript.Turns)
    Console.WriteLine($"{turn.Speaker}: {turn.Text}");

Models download on first use into a per-user cache, verified by SHA-256. The default clustering diarizer does not detect overlapping speech; an overlap-aware tier-2 diarizer is available by name. Read the measured results and limits before choosing a model.

Source and documentation · Project · Licences · Third-party notices

Container audio and subtitles

The optional RecursiveAI.OneSpeech.Media package adds OneMedia container audio input, Transcript subtitle exports (SRT, WebVTT, ASS and TTML), cue interchange and caption muxing. Usage and fidelity policy.

Command line

dotnet tool install -g RecursiveAI.OneSpeech.Cli provides onespeech transcribe, align, diarize, enroll and models pull.