Launching AimeFlux - Private, local MacOs dictation app.
AimeFlux is a Mac OS application designed to be private and local. It's secure by default. Your history, your settings, config, and all your data stays encrypted.
From MacOs 13.3 onwards, apple sillicon only. Tested with the base M1 Mac mini (16gb version) as a baseline.
No subcriptions, lifetime access, all future updates included.
Press a hotkey, speak, press it again, see it pasted.
Your clipboard remains intact by default unless you explicitly toggle the option to have them remain on it. Support for Universal Clipboard built in too.
Available Features
All the features that you'd expect from a dictation app these days and some workflow-related features based on my own usage and things I wanted to have without paying for subcriptions.
Local and private by design
You get Local processing using Whisper models. I use OpenAI's Whisper tiny, small, medium, and large models for multilingual workflows, and I also provide English-only versions based on Distil-Whisper for those that don't require all the language options and would prefer a faster experience.
Based on my experience, and taking into acount I'm not a native english-speaker, the english only variants are very fast and very good at taking my dictation even with my accent.
Support for optional LLM cleanup is built in. It's set up so that you can use your own API keys for the service or provider of your preference. But we also support local models like Ollama, and you can also set up your own providers. No information ever has to leave your machine if you don't want to.
History and Reprocessing
You get a full history UI for all your past transcriptions, dictations, and text imports. Your history remains local in an encrypted database and can be exported at any given moment if you want to. History items are immutable, so you're able to delete but not modify them. And if you ever want to change or reuse an item in your history, you can reprocess it.
Reprocessing gives you the flexibility of reusing the same raw transcription (text, not voice recording) and running it through different cleanup prompts, different replacements, and rules. This can be usefull if you dictated something in the wrong mode or want to reuse something you've already said in a different way (say, you said what you were going to work on on a standup message in slack and now want to format that as a task or issue for GitHub or an agent). You can always inspect or compare what changed in your transcript in the history.
Your voice and what changes were applied to it shouldn't be a black box, and this isn't it.
File Transcriptions
You can have one file or batch file transcriptions with optional cleanup, with mode rules, and SRT, VTT and Json export formats.
You can also set up watched folders so that you get automatic transcriptions on anything that comes into that folder. Transcriptions work with audio and video files with many supported formats, and you may enable or disabled a watched folder processing at any time without deleting it too.
Vocabulary and Replacements
You can set up Global Rules for vocabulary and common replacements for names, words and things the model might misshear. But you can also scope those vocabulary rules into modes so that techincal or work-related terms don't pollute your day to day dictation.
When Mode vocabulary is set, it replaces the global vocabulary when that given mode is active.
When Mode replacements are set, the pipeline applies all global replacements first, then applies your per-mode replacements next. You can see in the history detail what replacements were applied at any point and where the rules came from (globally or mode).
MCP Server and CLI
AimeFlux chips with a built-in MCP server and a fully functional CLI that allow you to extend how you work with it.
The MCP server exposes some tools for local agents to enable them to search through your transcriptions and history without you having to manually search for text and click it. It can also give you stats into how you use the tool or reprocess history items in a different context, or you can simply use it to fetch from the history some things you dictated a couple of days ago and work on it if you want to. Your history is available to it at any moment.
The CLI allows you to do almost everything you can do on the app and the UI, but without having to rely on it. It allows for more complex workflows, setups, and is a good entry point for automations.
Both the MCP server and CLI don't require the app to be running to work. Only dictation and Watched folders processing require the app to run. One-off processing can still be done, though
Settings Export and Import
All Modes, settings and preferences can be set up once and exported if you ever want to have the same config in another machine, to save you the trouble of configuring everything from scratch. The only thing AimeFlux doesn't export are your secrets, so you'll be prompted to add them again on import (like API keys).
I have been working on this for a couple of months. This is where I came to after a lot of back and forth. This has been a project that I've really liked, and I'm open for suggestions or changes. The current released version doesn't have auto updates or in-app updates yet, but I'm planning on adding that at a later point. I also offer a 21-day trial period with all features unlocked. And as I said before, this is a lifetime license. There are no subscriptions. And I don't plan on releasing another major version or something that will require you to pay for another app. All my updates will be on this one.
For lauch week, there is a 20% off deal on all licenses with code AIMELAUNCH20 until the 26th next week.
So you can get it for as low as 20$ for a single device.
Site: https://aimeflux.com/
BTW, I have a working Raycast extension pending approval that will give you some easy access to the history and other tools so you don't even need to open the UI for it, sent it yesterday for approval so might take some days before it's available in the store, I'll keep you posted!
I will try to set up a Discord server (haven't done it before) so I'm available for support or issues if you need!