AIAny
AI Audio2025
Icon for item

FluidVoice

On-device macOS dictation that transcribes speech locally and offers an optional local AI enhancement (Fluid Intelligence) for smart formatting and post-processing. Key features: low-latency model choices, live transcription overlay, per-app prompts and privacy-by-default; best on Apple Silicon.

Introduction

Most consumer dictation tools either send audio to the cloud for accuracy or sacrifice responsiveness to stay local. FluidVoice flips that trade-off by combining several low-latency STT engines with an optional, locally running enhancement model so you get near-instant insertion plus context-aware polishing without sending data off your Mac.

What Sets It Apart
  • Low-latency, multi-engine approach: supports Parakeet (near-instant English), Nemotron, Cohere, Apple Speech and Whisper. So what: pick a model based on latency vs. accuracy and swap quickly during onboarding.
  • Local AI enhancement (Fluid Intelligence): a private, on-device post‑processing runtime trained on dictation data for smart capitalization, punctuation and formatting. So what: transcripts look like edited text immediately, not raw ASR output, while audio and text remain local.
  • Live preview + system integration: notch-aware overlay, smart typing via accessibility APIs and per-app prompt configuration. So what: you can dictate directly into any app with contextual behavior per target app.
  • Privacy-first defaults: analytics are opt-in and core flows keep audio/text on-device. So what: suitable for sensitive workflows where cloud transcription is unacceptable.
Who It's For & Trade-offs

Great fit if you need private, fast dictation on macOS (writers, accessibility users, power note-takers) and want optional local AI polish without cloud dependencies. Look elsewhere if you need a fully cross-platform, server-hosted transcription fleet, or if you only have Intel macs and require the highest-accuracy large Whisper models (those run but with higher resource cost). Also note Fluid Intelligence is currently a private, optional runtime with a ~3.5 GB model footprint and RAM requirements when active.

Where It Fits

Positioned between built-in OS speech (zero-download but limited formatting) and cloud-hosted services (high accuracy, remote processing). FluidVoice aims to deliver near-cloud convenience with local privacy and near-real-time responsiveness on modern Macs.

Information

  • Websitegithub.com
  • Organizationsaltic-dev
  • Published date2025/09/21

Categories

More Items

Hugging Face
AI Audio2026

Performs speaker diarization (who spoke when) for live and recorded audio using an open-weight, 100M-parameter streaming-capable model that supports up to eight anonymous speaker channels, overlapping speech, chunked processing, and configurable latency for ASR integration.

Hugging Face
AI Audio2026

Performs low-latency streaming speech-to-text, emitting one token per selectable 80/120/160 ms clock with configurable transcription delay and a 30s rolling KV cache for unlimited 24/7 transcription. Bilingual (zh/en) and includes semantic VAD.

Hugging Face
AI Audio2026

Provides low-latency, true-streaming automatic speech recognition that emits append-only committed transcripts to avoid partial-result rollbacks. Supports configurable decoding chunks (80 ms–2 s), optimized for Chinese and English, and offers vLLM and transformers backends for real-time deployment.