UI improvements throughout, including better clean up and indicators when third party models are being loaded.

- Improved the clean up UI to be a bit more intuitive. - Added model picker to main view - Improved reliability converting video files

Improved loading the Parakeet and Whisper models to prevent cases where the transcription would not start.

Added Whisper from OpenAI as a new model option. This model processes slower, but is the most accurate local model available. Also made an improvement to the sidebar to look a little nicer.

Added VTT subtitle export support.

Quick Subtitles 2.0 is a major update with a redesigned experience and powerful new features. - Multi-Document Support - All your transcriptions are now saved and organized so you can go back to them anytime - Editable Transcripts - Tap to edit any part of your transcription directly - Clean Up - Find and fix transcription errors with suggested corrections you can review inline (or use YOLO mode to auto-apply them all) - NVIDIA Parakeet Engine - A new speech engine option that's up to 2x faster on Apple Silicon - New App Icon Plus numerous bug fixes and performance improvements.

- Improved UI consistency in live activity - Added an option to sign up for a subscription to use the AI features (bring your own key is still here, nothing is exclusive to the subscription). This aims to give a solution to people who don't want to manage their own API keys - Improved benchmark history design - Improved batch processing design

Live activity and background transcription support is here! Start a transcript and go do other things while it works in the background. - Improved line break logic for SRT exports - Added a benchmark download to the benchmark history page.

Benchmark mode has a history section with basic charts and downloadable data. The UI has been refreshed a bit for the new year.

Added a benchmark mode for premium users (find it in settings). GPT-5.2 added for the analysis feature.

- Added support for Gemini 3 Flash - AI Analysis features render responses as rich text, not raw Markdown

Feature update: Set a default save directory so all subtitle files save to the same place every time Thank you for using Quick Subtitles!

Several OpenAI API updates: - Model picker, supporting the latest GPT-5 models (5.1, mini, and nano) - Ability to verify your OpenAI API key Gemini has better response text when something goes wrong.

1. Added the option to use your own OpenAI API key (this API is much slower than Gemini, so be patient if you use it!). 2. Added option to customize the the analysis prompt to whatever you want.

Working on fixing the issue preventing loading the in-app purchase.