--- title: "onespeech docs" url: https://recursiveai.org/onespeech/docs/ description: "Everything about OneSpeech 1.0.0: getting started, transcribing files, live sessions, word timing, diarization and speaker recognition, the model…" version: "1.0.0" --- # onespeech docs Everything about OneSpeech 1.0.0: getting started, transcribing files, live sessions, word timing, diarization and speaker recognition, the model families, model files and offline packages, captions, and every measured result. These docs are for OneSpeech 1.0.0, read at the v1.0.0 tag (63002de). ## Start - [Getting started](https://recursiveai.org/onespeech/docs/getting-started/index.md): Install OneSpeech for .NET 10, transcribe a recording with timed words and speaker turns, run a live session from packaged models with no network, and use… - [Overview](https://recursiveai.org/onespeech/docs/overview/index.md): OneSpeech transcribes audio, times words, labels speakers and supports live sessions in managed .NET 10. Open-weight models run on OneModel; OneAudio… - [Transcribing files](https://recursiveai.org/onespeech/docs/transcribing-files/index.md): Speech recognition for .NET. OneSpeech is unaffiliated with other products or companies named OneSpeech. - [Live sessions and file streaming](https://recursiveai.org/onespeech/docs/live-sessions/index.md): A LiveSession accepts borrowed mono 16 kHz AudioBuffer blocks through synchronous Push, or awaits those blocks through Listen. Resampling and capture… - [Shared contracts](https://recursiveai.org/onespeech/docs/contracts/index.md): AudioBuffer views interleaved float PCM without copying it. Mono is one channel; multiple channels preserve separated-party audio for later channel… - [Transcript JSON, version 1](https://recursiveai.org/onespeech/docs/transcript-json/index.md): UTF-8 JSON, written and read using a generated System.Text.Json context. The root object has required schemaVersion (integer major version, currently 1)… ## Coming from another library - [From Whisper.net](https://recursiveai.org/onespeech/docs/migrate-whisper-net/index.md): This is a concept map, not an API-compatible shim. It uses the published package documentation, consulted 2026-11-08; no implementation source is needed. - [From sherpa-onnx .NET](https://recursiveai.org/onespeech/docs/migrate-sherpa-onnx/index.md): This guide maps concepts from the published C# documentation, consulted 2026-11-08. It is not a shim and does not import model configuration objects or… - [From Azure Speech SDK](https://recursiveai.org/onespeech/docs/migrate-azure-speech/index.md): Here “offline recognition” means recognizing a file rather than a microphone; it does not imply that an Azure service call itself is offline. This concept… - [From WhisperX](https://recursiveai.org/onespeech/docs/migrate-whisperx/index.md): This guide uses only the published package description, consulted 2026-11-08. It maps a workflow, not Python modules, dictionaries or an API signature. ## Audio, timing and biasing - [Audio input and the pipeline](https://recursiveai.org/onespeech/docs/audio-input/index.md): AudioFileSource is a synchronous pull decoder. It accepts seekable encoded streams or paths and returns interleaved float PCM at 16,000 Hz, preserving the… - [Long files and cancellation](https://recursiveai.org/onespeech/docs/long-files/index.md): Use Speech.Transcribe when you need a complete document with whole-recording speaker labels. Use Speech.TranscribeStream when you want final segments as… - [Voice activity detection](https://recursiveai.org/onespeech/docs/vad/index.md): The core assembly embeds the unmodified silero_vad.onnx from Silero VAD v6.2.3, revision 5cd7945676eb32225748052e2e6a0580e4686a08, SHA-256… - [Word timing and alignment](https://recursiveai.org/onespeech/docs/word-timing/index.md): Every Word has nullable Start and End, nullable Confidence, and TimingMethod. Times are TimeSpan values on the original audio clock, not processing time.… - [The CTC alignment trellis](https://recursiveai.org/onespeech/docs/ctc-alignment/index.md): CtcTrellis.Align accepts row-major natural-log frame probabilities and a fixed sequence of nonblank vocabulary IDs. It consumes every frame and requires… - [Contextual phrase biasing](https://recursiveai.org/onespeech/docs/biasing/index.md): Contextual biasing favours domain names and phrases during recognition; it is not model training and cannot guarantee that a phrase is spoken. The… - [Contextual token scoring](https://recursiveai.org/onespeech/docs/contextual-biasing/index.md): OneSpeech.Decoding.PhraseTrie is an immutable, inference-only context graph. Construct it from TokenPhrase values tokenized by the model's tokenizer.… ## Speakers - [Diarization](https://recursiveai.org/onespeech/docs/diarization/index.md): Speech recognition for .NET. Speaker labels are document-local, not biometric identities. No audio or embedding is persisted by diarization helpers. - [Live speaker labels](https://recursiveai.org/onespeech/docs/online-diarization/index.md): Speech.Listen adds bounded online clustering to live recognition. The low-level OnlineDiarizationSession decorates a fresh recognition-only LiveSession… - [Nemotron 3 Diarization](https://recursiveai.org/onespeech/docs/nemotron-diarization/index.md): Speech recognition for .NET. OneSpeech.Diarization.NemotronDiarizer is an explicitly selected, overlap-aware, tier-2 file diarizer for up to eight… - [Gated models and segmentation](https://recursiveai.org/onespeech/docs/gated-segmentation/index.md): Gated entries are tier 2, explicitly named, never defaults or mirrors. The catalog pins publisher revisions and each file's SHA-256. ModelStore.PullAsync… - [Speaker recognition](https://recursiveai.org/onespeech/docs/speaker-recognition/index.md): Diarization asks who spoke when, using anonymous labels. Recognition asks whether speech matches an enrolled identity. InMemorySpeakerRegistry borrows a… - [Speaker identity and scores](https://recursiveai.org/onespeech/docs/speakers/index.md): SpeakerEmbedder is the synchronous CPU extraction role. Its model metadata identifies the checkpoint and licence; Dimension fixes output length and… ## Model families - [Whisper](https://recursiveai.org/onespeech/docs/whisper-features/index.md): OneSpeech.Audio.WhisperLogMelSpectrogram consumes at most 480,000 samples of 16 kHz mono float PCM. Resampling and channel mixing belong to the audio… - [Qwen3-ASR and forced alignment](https://recursiveai.org/onespeech/docs/qwen3/index.md): Qwen3-ASR-0.6B and Qwen3-ASR-1.7B run on OneModel, including their AuT audio encoder/projector and Qwen language decoder with a past/present KV cache. The… - [Moonshine](https://recursiveai.org/onespeech/docs/moonshine/index.md): The registered English checkpoints are moonshine-ai/moonshine-tiny, moonshine-base, moonshine-streaming-tiny, moonshine-streaming-small and… - [NeMo and transducers](https://recursiveai.org/onespeech/docs/nemo-and-transducers/index.md): NemoLogMelSpectrogram consumes mono 16 kHz float PCM. It implements the Parakeet v3 preprocessor parameters: 128 Slaney-area-normalized mel bands… - [Cache-aware FastConformer](https://recursiveai.org/onespeech/docs/cache-aware-conformer/index.md): Nemotron 3.5 ASR loads only by explicit model name, nvidia/nemotron-3.5-asr-streaming-0.6b. It is an upstream-only OpenMDW-1.1 model, never a default or a… - [Custom speech graphs](https://recursiveai.org/onespeech/docs/custom-models/index.md): OneSpeech.Custom.GraphSpeechRecognizer is a low-level SpeechRecognizer for caller-supplied float32 ONNX or OneModel files. It never downloads or converts… - [Integer inference](https://recursiveai.org/onespeech/docs/int8/index.md): The fp32 defaults are unchanged. The following explicit catalog IDs select locally converted, mixed-int8 graphs: ## Models and files - [First-use models](https://recursiveai.org/onespeech/docs/model-downloads/index.md): model selects the requested role, not an isolated inference graph. Present files are verified and reused. Nothing is downloaded on Transcript export.… - [Model files and pre-fetch](https://recursiveai.org/onespeech/docs/models/index.md): ModelCatalog.Default reads a versioned embedded data catalog without network access. ModelStore.PullAsync(id) resolves an explicit model ID to a… - [Loading and composing roles](https://recursiveai.org/onespeech/docs/loading/index.md): Every downloadable family has the same canonical loader pair: - [Tier-2 models and gates](https://recursiveai.org/onespeech/docs/tier-2-models/index.md): Tier 2 includes CC-BY-4.0, OpenMDW-1.1 and every gated model even if its licence is MIT. Select these models explicitly by catalog ID. They are never… - [Command line and model packages](https://recursiveai.org/onespeech/docs/cli-and-model-packages/index.md): Speech recognition for .NET. onespeech --help lists commands, positional arguments, options and exit codes. The tool accepts standalone audio and OneMedia… - [Containers and captions](https://recursiveai.org/onespeech/docs/media/index.md): RecursiveAI.OneSpeech.Media references core, OneMedia and OneMedia.Audio. Call OneSpeech.Media.MediaInput.Register() once during startup, before… ## Measured results - [Measured results and limits](https://recursiveai.org/onespeech/docs/results/index.md): The release revalidation table distinguishes full-split measurements from the earlier fixed-slice evidence below. The tables disclose incomplete datasets… - [Evaluation](https://recursiveai.org/onespeech/docs/eval/index.md): Speech recognition for .NET. These tools score every system with the same reference text and normalizer. They are not part of the runtime package. The… ## Inside the library - [Model-data and document readers](https://recursiveai.org/onespeech/docs/model-data-readers/index.md): This is the inventory of file-reading code owned by OneSpeech. The mutation targets are development tools; measured results report the completed campaigns… - [The safetensors reader](https://recursiveai.org/onespeech/docs/safetensors-reader/index.md): The internal OneSpeech.Formats.SafeTensorFile reads the published safetensors format (v0.6.2 format description). It does not load executable objects. A… - [Third-party notices](https://recursiveai.org/onespeech/docs/third-party-notices/index.md): Core embeds the Silero VAD v6.2.3 ONNX release model. No other model weights are included. - [Changelog](https://recursiveai.org/onespeech/docs/changelog/index.md): All notable changes are documented here, following Keep a Changelog.