Changelog

Release notes for VivaDicta for Mac.

1.2.0

April 20, 2026

A release focused on making VivaDicta scriptable and agent-driven. A new command-line tool, an expanded MCP server, and an official agent skill pack.

Command-line tool

  • New vivadicta CLI installable via brew install n0an/tap/vivadicta
  • Browse history, search with filters, export to Obsidian, rewrite with any preset, transcribe files or YouTube URLs — all from your terminal
  • TTY-aware output: readable tables in your shell, JSON when piped into jq, a script, or a Raycast action
  • Shell completions for zsh, bash, and fish via vivadicta --generate-completion-script

MCP server

  • VivaDicta now exposes 12 tools to AI assistants (Claude Code, Claude Desktop, Cursor, Codex CLI, Zed, VS Code, and any other stdio MCP client)
  • Reads: search and fetch transcriptions with natural-language timeframes; list presets, tags, Viva Modes, vocabulary, word replacements
  • Writes: rewrite with any preset; transcribe local audio files or YouTube URLs with live progress polling; poll async job status
  • Install via brew install n0an/tap/vivadicta-mcp

Agent skill pack

  • Official skill pack so AI assistants know when and how to call vivadicta for you
  • Install with npx skills add n0an/vivadicta-cli-skills --global
  • Covers transcribe, search-and-rewrite, and vault-export flows end-to-end

Clipboard setting fix

  • The toggle in Settings → Clipboard was labeled "Copy to Clipboard" but, when on, actually disabled clipboard restoration
  • Renamed to "Restore Clipboard After Paste" so the label, behavior, and the nested "Restore Delay" picker all agree
  • Existing preference is migrated automatically

Provider polish

  • VivAgents server surfaces a clear error in the UI when the port is already in use, and exposes the port it's actually listening on
  • Custom OpenAI-compatible providers auto-append /chat/completions when only the base URL is pasted
  • Cerebras default model updated to llama3.1-8b
  • MLX generation parameters tuned to reduce repetition in outputs

1.1.0

March 28, 2026

Major update with sign-in support, new translation presets, local AI models, tags, export, and YouTube transcription.

New Translation Presets

  • 8 new built-in translation presets: Chinese, French, German, Portuguese, Japanese, Korean, Arabic, Italian

New transcription provider

  • Cohere Transcribe — best-in-class accuracy (#1 on Open ASR Leaderboard), 14 languages, free trial tier

Sign in with your subscriptions

  • GitHub Copilot — sign in and access Claude, GPT, Gemini, Grok (free tier available)
  • ChatGPT — sign in with your ChatGPT Plus/Pro subscription, no API key needed
  • Google Gemini — sign in with your Google account for AI processing
  • OpenRouter — one-click sign-in, API key generated automatically
  • VivAgents Server — use CLI agents (Claude, Codex, Gemini) locally or share with iOS devices

Local AI with MLX

  • Run open-source LLMs locally on Apple Silicon — free, private, no internet
  • Qwen 3.5 models (4B, 9B, 27B, 35B-A3B MoE)
  • OpenAI GPT OSS 20B, Ministral 3B, Ministral 8B
  • Model browser organized by Mac RAM tier with recommendations
  • Base, Med, High quantization variants

Tags & Organization

  • Custom tags with name, color, and icon
  • Source tags auto-track origin: Mac, iOS, iOS Keyboard, Share Extension
  • Filter transcription history by source and custom tags
  • Tags sync across devices via iCloud

Export

  • Export transcriptions to CSV, JSON, or Markdown
  • Bulk export from history or individual export from detail view
  • Markdown export includes metadata header

YouTube Transcription

  • Transcribe YouTube videos by pasting a URL
  • Optional timestamps
  • AI processing available on transcribed content

1.0.0

February 27, 2026

Initial release of VivaDicta for Mac.

  • System-wide voice-to-text — works in any app via global shortcut
  • Local transcription with Whisper, Parakeet, and Apple Speech
  • Cloud transcription via Groq, Cohere, Deepgram, ElevenLabs, Gemini, Mistral, Soniox, and more
  • AI text processing with 39 built-in presets (Anthropic, OpenAI, Gemini, Groq, Ollama, and more)
  • Viva Modes — auto-configure model, language, and AI preset per app or website
  • Transcription history with search, export, and audio playback
  • iCloud sync with VivaDicta for iOS
  • Custom vocabulary and word replacements
  • Mini recorder with keyboard shortcuts for fast preset switching