Launched this week
Seminay
Minimalist Windows Voice Assistant with Gemini Live API
5 followers
Minimalist Windows Voice Assistant with Gemini Live API
5 followers
Seminay is a simple, lightweight voice assistant for Windows that works with the Gemini Live API. Features: Voice Chat: Speak directly to Gemini and hear its response in real time. System Tray: It runs quietly in your Windows system tray. Open Source: Built with Python
Greetings Everyone,
I wanted to share the interface I built for myself, inspired by the AI "Samantha" from the movie "Her," after making a few revisions to it.
I tried to keep it as simple and user-friendly as possible. Just get a free API Key from "https://aistudio.google.com", enter it into the prompt when you first launch the app, and you're good to go.
If you wish, you can change it later by entering a different key in the settings.
You can start the conversation either by speaking into the microphone right away or by double-clicking the capsule to type your message in the text field that appears. Gemini will only respond via voice.
As I mentioned, I tried to keep it quite minimal. Features like screen capture, camera, etc., have been removed. It is strictly for voice-based responses and chatting. (However, they might be added in the future if requested.)
I am not someone who knows how to code; I had AI write the code entirely, so please overlook any flaws. As far as I know, "Gemini Live" is an emerging technology, and we are using "api_version": "v1alpha" in the code.
Unfortunately, I couldn't get the other models to work. Due to these reasons, token limits, or server-side issues, there might be occasional interruptions. If it disconnects, try restarting the app at the very least; it will likely fix the issue.
Anyone is welcome to modify, improve, or take inspiration from this app for their own projects however they see fit—it's entirely up to you... Much love to everyone...
---
Seminay AI v.0.0.1 - Update Notes
I believe I have resolved the connection, memory, and Google Search issues. I personally tested it for 4.5 hours; as long as the application remains open, Google Search is active, Gemini can access up-to-date information, the connection refreshes, there is no noticeable interruption, and the memory remains intact... (Note: Due to Google authorization token limitations, Gemini may occasionally lock the interface in "thinking" mode when a query requires a search. If this happens, simply speak to it again; this will refresh the connection and search limits, allowing you to continue the conversation from where you left off.)
The data transmission method has been updated to the current method required by Google.
Various code and interface improvements were made. (Transparent background, dragging, shadow effects, etc.)
honestly this looks really handy, love that it just sits in the tray. one thing though, can you add a hotkey to push to talk? like holding a key to record instead of clicking would make it way easier when im already typing in another window
@lknurbykgesmo2 "Thank you for your interest and feedback. First of all, let me clarify that in the current version, the microphone is on by default; even when the app is hidden in the system tray and you are focused on other tasks, it captures your voice and responds. Google offers an uninterrupted tunnel for this service. However, from what I understand, you want your voice to be transmitted only when you choose to—essentially a 'walkie-talkie' logic—and have Gemini not receive your voice at all otherwise... (This actually means sending less data over the constantly open channel, which would make the app more stable, consume less bandwidth, and prevent accidental Gemini triggers from background noise—making your suggestion highly logical...)
In fact, I had initially considered this, but due to issues like 'Permission and Conflict (UAC) Troubles', 'PyQt and Asynchronous Event Loop Conflicts', 'Key Conflicts with Other Applications', etc., I decided not to add any global hotkeys except for the ESC key. So, I am actually well aware of the situation; I might have taken a bit of a shortcut to avoid potential headaches, yes.
However, I have received and noted your feedback. I will work on how we can make this feature more usable in the most stable way possible. I'm already working on other upcoming features anyway, so I'll try to blend them all together. Thank you again for your interest and feedback."
love how minimal this is, feels like the gemini live api deserves a nice lightweight wrapper. one thing i keep wishing for is a global keyboard shortcut so i can summon the chat from anywhere without hunting for the tray icon. a small popup window with voice capture would turn it into something close to an always on assistant
@nevzatsaymnaqn "Dear Mr. Nevzat, thank you for your interest and feedback, I'm glad you liked it! As I mentioned in my reply to Ms. Ilknur; 'In the current version, the microphone is on by default; even when the app is hidden in the system tray and you are focused on other tasks, it captures your voice and responds.' In other words, as long as the microphone is active (the green light on it is on), it is active under all conditions.
However, from what I understand from your words, you want to bring the user interface back from its hidden state. Meaning, you are saying it would be great to summon the 'Show' button with a shortcut instead of using the system tray. Yes, this is definitely possible, and this feature actually existed before. But as I explained in my other message, since global hotkeys are quite problematic, I took a bit of a shortcut and left it out. Nevertheless, I have noted your request; I will work on this and other updates to make the app more stable and user-friendly. Thank you again for your interest and feedback.
(NOTE: I would also like to remind you that this application is completely open-source and free. Experienced developers can certainly take this foundation I built and turn it into something much more professional. I would be more than happy about that, but I will also continue to work on improving it myself.)"