Meetily Pro 1.1.1 - The New Beginning Release Notes

What's New in Pro 1.1.1

The biggest Meetily update yet. This release rebuilds the transcription engine from the ground up with pluggable providers, adds support for 98 languages, fixes the transcript loss issue that affected long meetings, and introduces professional editing tools.

233 commits, 25 PRs merged, 184 files changed. 19 days of development.

98-Language Support

Full-stack multi-language support across transcription, summaries, and PDF export.

Inline language selector - Pick your language from a combobox right in the recording view. No more navigating to settings
Language-aware summaries - Summaries now generate in the same language as your transcript. German meetings get German summaries, French meetings get French summaries
CJK and complex script support - Chinese, Japanese, Korean, Thai, Lao, Khmer, and Myanmar scripts render correctly in both transcripts and PDF exports
Smarter Whisper prompting - Non-English languages skip the English initial prompt, improving accuracy for languages like Swiss German, Finnish, and Vietnamese

This was the most requested feature since January, with users across Germany, Switzerland, France, Belgium, Finland, and Southeast Asia waiting for native language support.

Pluggable Transcription Architecture

Meetily is no longer limited to built-in models. Connect any transcription backend you want.

Provider	Type	How it works
Whisper	Built-in, local	GPU-accelerated on Metal (macOS), CUDA, and Vulkan (Windows)
Parakeet	Built-in, local	NVIDIA's streaming ASR engine for Windows
OpenAI-compatible	External API	Connect Groq, Speaches, Together, Fireworks, or any OpenAI-compatible endpoint
HuggingFace custom	External models	Download and run any GGML Whisper model from HuggingFace

9 provider presets included - Select a provider, enter your API key, and you're ready. Meetily handles connection testing, model discovery via /v1/models, and prompt configuration automatically.

Per-provider settings - API keys and server URLs are stored separately for each provider. Switching between Groq and your local Speaches instance won't mix up credentials.

HuggingFace Custom Models

Download any GGML Whisper model directly from HuggingFace with full validation.

Browse and select - Enter a HuggingFace URL, browse available GGML files in the repo
Smart validation - 6-layer format checking catches incompatible models (GGUF, F32, PyTorch .bin) within the first 48 bytes, before you waste bandwidth on a multi-GB download
Cancel and resume - Cancel in-progress downloads. Partial files are cleaned up automatically
Gated repo support - Public repos work without tokens. Gated repos prompt for your HuggingFace token

Transcript Reliability Fix

The most critical fix in this release. Previously, stopping a recording could lose the final segments of your transcript because the pipeline exited before the transcription queue finished processing. For long meetings, this meant losing minutes of content.

What changed:

Reordered stop sequence - The transcription pipeline now fully drains before the recording flag clears
Whisper prompt context bleeding fix - Words from your previous meeting no longer contaminate the next recording
4-stage post-processing pipeline - Every transcript segment now passes through confidence scoring, hallucination filtering, words-per-second analysis, and deduplication

This fix addresses the issue reported by multiple users where meetings would record full audio but only produce partial transcripts.

Transcript Editing

Professional editing tools for your transcripts.

Inline editing - Click any text to edit it in-place. Changes auto-save
Find & Replace - Press Cmd/Ctrl+H to search and replace across your entire transcript. Works in the summary editor too, including inside tables
Click-to-seek - Click any timestamp to jump audio playback to that exact moment
Smart click zones - Clicking a timestamp seeks audio. Clicking text opens the editor. No more accidental mode switches

Audio Streaming Server

Large recordings (1hr+) now play instantly. We replaced the old in-memory approach with a local streaming server that serves audio with HTTP Range requests.

Low memory - ~8KB per request instead of loading the entire file into memory
Instant playback - Seek to any point in a long recording without buffering
Secure - Bound to localhost only, serves only audio files from your recordings directory, with path traversal protection

Platform Notifications

Custom notification systems on both platforms, replacing generic OS notifications.

macOS - Liquid Glass toast with vibrancy effects, Siri-glow animation, and dynamic app dropdown showing detected meeting apps. Multi-display support, full dark/light mode.

Windows - Custom borderless always-on-top notification with Liquid Glass UI, app badge display, and dark/light mode. Falls back to WinRT toast if needed.

Multi-Language PDF Export

Export your transcripts and summaries to PDF in any of the 98 supported languages.

Native save dialog with Reveal in Finder/Explorer after export
CJK character rendering with embedded fonts
Complex script support - Thai, Lao, Khmer, Myanmar with proper hyphenation
Nested sublists render correctly in exported PDFs

Summary Model Chooser

Improved model selection for the summary AI engine.

Visual model chooser - Browse and select your preferred summary model
Remembers your choice - Auto-enabled after your 3rd view
Credit tracking - See your remaining Hosted AI credits at a glance
Instant display - Credits are cached at startup so the count shows immediately
Model deletion guard - Can't accidentally delete the model you're actively using

Download Resilience

Model downloads are now more reliable.

Resume on interruption - Network drops during a model download? It picks up where it left off
CDN fallback - If the primary download source fails, Meetily automatically tries an alternative
Progress fix - No more crashes from download progress exceeding 100%
Concurrent download handling - Starting a second download while one is in progress no longer triggers false error banners
Pre-flight mic check - macOS now verifies microphone permissions before you start recording

Bug Fixes

Fix	Impact
Prevent transcript data loss during post-stop worker drain	Critical
UTF-8 panic in Whisper dynamic prompting for CJK text	Critical
Whisper prompt context bleeding between recordings	High
Large audio file playback failure	High
BlockNote summary editor infinite recursion	Medium
Model chooser dropdown hidden behind sidebar	Medium
Recovery dialog and alerts hidden behind overlays	Medium
Stale API key persistence across provider switches	Medium
Audio player not loading on Windows production	Medium
Ollama null model list crash	Medium

Technical Details

Component	Change
Version	0.1.16 → 1.1.1
Development period	19 days (Feb 19 - Mar 10)
Commits	233
PRs merged	25
Files changed	184
Lines added	34,711
Lines removed	5,886
New Rust modules	8
New React components	5
Languages supported	98
Transcription providers	4 (Whisper, Parakeet, OpenAI-compatible, HuggingFace)
Provider presets	9
New Tauri commands	13
Unit tests (lib)	274 pass
KPI integration tests	20/20 pass
Contributors	3

Upgrade

Seamless upgrade from v0.1.16. All existing settings, recordings, and models are preserved. New features activate automatically.

Download

macOS (Apple Silicon): Download .dmg(opens in new tab)
Windows: Download .exe(opens in new tab)

Release Highlights