Hex: local voice dictation that pastes into the focused desktop app
Hex is an open-source dictation app that transcribes locally and pastes into the focused app; download its model once, then dictate on Mac, Windows or Linux.
Published entries across all sections carrying the “Speech to text” tag, newest first by publication date on this site.
14 entries
Hex is an open-source dictation app that transcribes locally and pastes into the focused app; download its model once, then dictate on Mac, Windows or Linux.
An Apache-2.0 meeting assistant: live bilingual subtitles, structured notes, push-to-talk interpreting. Four platforms, Mac on-device. 1.4k stars.
Open-source Python framework by Daily: speech, TTS and multimodal pipelines are plug-and-play, multi-agent handoff is built in, SDKs cover seven platforms.
AudioScript listens to system audio or your mic, transcribes and translates locally with sherpa-onnx and opus-mt, and overlays English-Chinese subtitles.
earshot is an open-source Python toolkit that lets AI agents hear: describe sounds, score fit, find best moments, spot synthetics, make sound in code.
Cactus Compute open-sourced Needle: an 8-29 MB on-device model for tool calls, extraction and speech recognition, phones to microcontrollers, Apache-2.0.
VoxWeft ports VoxCPM2 to MLX for on-device interpretation: first audio in about 170 ms, 30 languages, 5-second voice cloning, nothing leaves the machine.
Ming 3.0 TTS, ASR and audio understanding: 0.5B and 7.9B TTS variants, context-aware transcription, and a 126B audio model with demanding GPU needs.
Google AI Edge Foresight is Google free experimental Mac notetaker, launched October 6: EmbeddingGemma 2 and Gemma 4 transcribe meetings on-device, offline.
Self-hosted meeting middleware. Agents join browser video calls over MCP, read the live transcript and speak or chat, with your choice of LLM and voice.
Adds two-way voice conversations to Claude Code and other agents. Runs offline with local Whisper and Kokoro, or falls back to OpenAI cloud speech services.
MAI-Transcribe-2-Streaming takes No.1 on Artificial Analysis for final and partial transcripts: 2.50% WER, 0.13s latency, 60 languages, $0.54/hour promotional.
AutoClip is an open-source video clipper for long-form material: feed it an interview, podcast, lecture or gaming VOD and it mines the subtitles for an outline, scores segments for highlight-worthiness and cuts clips you approve before anything renders. Unlike cloud clipping services, cutting and rendering stay on your machine with your own model keys — fully offline via Ollama. It ships as a desktop app, a Docker stack and a CLI/MCP pipeline, with export presets for Douyin, Shorts and Bilibili.
AudioNotes runs speech recognition locally with FunASR, turning transcripts into Markdown via a local Ollama model, keeping data on-device for meetings.