# Fluid Inference

## Docs

- [Custom Pronunciation](https://docs.fluidinference.com/asr/custom-pronunciation.md): Override TTS pronunciation with custom lexicon files.
- [Custom Vocabulary](https://docs.fluidinference.com/asr/custom-vocabulary.md): CTC-based vocabulary boosting for domain-specific terms without retraining.
- [ASR Getting Started](https://docs.fluidinference.com/asr/getting-started.md): Batch and streaming transcription with Parakeet models.
- [Manual Model Loading](https://docs.fluidinference.com/asr/manual-model-loading.md): Deploy ASR models offline without HuggingFace downloads.
- [Streaming ASR](https://docs.fluidinference.com/asr/streaming.md): Real-time streaming transcription with Parakeet EOU and end-of-utterance detection.
- [Configuration](https://docs.fluidinference.com/configuration.md): Model registry, proxy settings, and environment configuration.
- [Diarization Getting Started](https://docs.fluidinference.com/diarization/getting-started.md): Speaker diarization — identify who spoke when in audio.
- [Offline Pipeline](https://docs.fluidinference.com/diarization/offline-pipeline.md): Full VBx batch diarization with pyannote-compatible pipeline.
- [Sortformer](https://docs.fluidinference.com/diarization/sortformer.md): NVIDIA's end-to-end streaming speaker diarization model.
- [SpeakerManager API](https://docs.fluidinference.com/diarization/speaker-manager.md): Track and manage speaker identities across audio chunks.
- [Streaming Diarization](https://docs.fluidinference.com/diarization/streaming.md): Real-time speaker diarization for live audio streams.
- [Audio Conversion](https://docs.fluidinference.com/guides/audio-conversion.md): Convert any audio format to 16 kHz mono Float32 for FluidAudio pipelines.
- [Manual Model Loading](https://docs.fluidinference.com/guides/manual-model-loading.md): Deploy models offline without HuggingFace downloads.
- [Installation](https://docs.fluidinference.com/installation.md): Add FluidAudio to your Swift project.
- [Introduction](https://docs.fluidinference.com/introduction.md): Local audio AI for Apple devices — speech-to-text, speaker diarization, voice activity detection, and text-to-speech on the Neural Engine.
- [Converting Models](https://docs.fluidinference.com/mobius/converting-models.md): How to convert PyTorch models to CoreML using möbius — export, validate, quantize.
- [Getting Started](https://docs.fluidinference.com/mobius/getting-started.md): möbius — convert AI models for edge deployment on Apple Silicon, NPUs, and other accelerators.
- [Quickstart](https://docs.fluidinference.com/quickstart.md): Get up and running with FluidAudio in minutes.
- [API Reference](https://docs.fluidinference.com/reference/api.md): Complete API reference for FluidAudio.
- [Benchmarks](https://docs.fluidinference.com/reference/benchmarks.md): Performance benchmarks across all FluidAudio capabilities on Apple Silicon.
- [CLI Reference](https://docs.fluidinference.com/reference/cli.md): FluidAudio command-line interface for testing and benchmarking.
- [Models](https://docs.fluidinference.com/reference/models.md): CoreML model catalog and HuggingFace sources.
- [TTS Custom Pronunciation](https://docs.fluidinference.com/tts/custom-pronunciation.md): Override TTS pronunciation with custom lexicon dictionaries.
- [Kokoro TTS](https://docs.fluidinference.com/tts/kokoro.md): High-quality text-to-speech synthesis with 48 voices.
- [PocketTTS](https://docs.fluidinference.com/tts/pocket-tts.md): Autoregressive TTS with dynamic audio chunking and streaming output.
- [SSML Support](https://docs.fluidinference.com/tts/ssml.md): Control pronunciation with Speech Synthesis Markup Language tags.
- [VAD Getting Started](https://docs.fluidinference.com/vad/getting-started.md): Voice activity detection with Silero VAD v6 on CoreML.
- [Segmentation Config](https://docs.fluidinference.com/vad/segmentation-config.md): Tune VAD segmentation for your use case.
- [Streaming VAD](https://docs.fluidinference.com/vad/streaming.md): Real-time voice activity detection with event callbacks.

## OpenAPI Specs

- [openapi](https://docs.fluidinference.com/api-reference/openapi.json)