The latest, useful and creative AI features are just around the corner.
Sure! I can summarize this page for you as follows:• It talks about an artificial intelligence framework that works entirely locally.• A digital assistant powered by artificial intelligence that can clone your voice.• It will provide a smooth and uninterrupted experience across devices.• Ability to interact with applications by using the screen reader feature.• An AI-powered speech-to-text engine will be introduced.*Not actual artificial intelligence response. Provided as an example.An artificial intelligence that runs entirely locally Demirdelen AI works using the most common and advanced artificial intelligence models. The Gemma2 model with 27 billion parameters, which almost the best performance among publicly available models (for now), was chosen as the main model because it also has high performance in languages other than English. Not only that, depending on the intended use, it can also choose special models for fields such as chatting, coding, summarization and art. All the artificial intelligence models for it run completely locally within the borders of Gönyeli, in Northern Cyprus. In other words, it never uses a existing service such as ChatGPT to generate any response and does not send a single HTTP request abroad, except for requests from foreign countries.A digital assistant like no other By using these advanced artificial intelligence models, it makes everyday life easier in much more creative and useful ways, unlike other digital assistants. Everything ranging from processing commands between devices to smart home control. And all with a very familiar voice.Powered by AIThe assistant accurately understands and processes the commands or messages coming from users, using models specially trained in this field. It can also provide extra information about the status of the command or message in cases where a response is required. *For voice responses, it can also create realistic conversations using AI voice cloning techniques.Advanced commandsInstead of just setting an alarm, reminding of this & that, opening an application and making internet searches like other digital assistants, you can ask any question you wish and take action with commands freely and without censorship. It also supports advanced commands that are expected but won't usually be supported by popular digital assistants, such as "download this video using my phone", "open this game from my computer", "summarize this", "translate into this language", and more.Just like thisWhen the assistant is launched, it automatically starts listening for a command. The command can also be typed by using a keyboard. (optional)It detects the end of the conversation by monitoring changes in noise level. It then processes the audio recording and displays it briefly for checking.Once the command is processed, it is then sent to the AI model. *Models running locally may take a little longer to generate a response.Lastly, the model's response is displayed on the screen and read with a cloned or normal voice, depending on the users preference.*UI and UX elements are not final and may change in future iterations. *Images are provided as an example. Sequences are shortened and are just representative.Cross-platform usage & access Regardless of device type, most of what is covered here will work on all supported Android phones and Windows computers. Other devices may also have a limited web app that will be hosted later, in the coming months.A screenshot showing the desktop web application. Most artificial intelligence features such as chatting, summarizing and drawing will be used through the web browser. *A service application installed on the users computer is required in order to run commands interacting with the computer.Phone appThe application that will be developed for Android™ smartphones will provide access to these AI models anytime, anywhere over the internet. Providing AI features that enchance the mobile phone experience, while also featuring a live chatting with the models. *Some smaller models will be able run directly on the phone without the need for internet later on.Watch appIn addition, all watches that support Wear OS by Google will be able to quickly chat with the AI assistant without needing a phone, right on your wrist. The app will also feature a toned-down version of advanced commands like "Reply on my phone" and "Start on my phone" for example.AI-Powered Screen Reader Digital assistant features aside, Demirdelen AI can understand what is happening on the screen when summoned and respond according to its contents. It can interact with buttons, boxes and text if necessary, performing actions on behalf of the user.Save with AIWhile viewing posts can contain photos or videos on supported smartphones and social media platforms, the AI assistant can seamlessly download and save the content to the phone's gallery.Like with AIAgain, while viewing posts on supported smartphones and social media platforms, the AI assistant can be summoned to like the post or make a comment.AI-Powered repliesIf someone sends a message at an inconvenient time on supported smartphones and platforms, a reply briefly talking about the current status of the user such as "Sleeping, busy, outside, etc." can be sent using AI.AI-Powered Speech-to-Text Engine By making the voice typing engine available on its own outside of the digital assistant, it can transcribe speech into text accurately without being affected by differences such as personalized accents using advanced voice detection methods. For example, when the artificial intelligence assistant is summoned while a text box is selected on the phone, it can optionally write the detected sentence directly into the selected box on behalf of the user.One more thing...AI-Powered Image Generation Lastly, Demirdelen AI will allow a small group of users who have allowed the use of their facial data to imagine themselves in different scenarios. The users facial data will not be shared with any third party, like for the use of advertising purposes like other AI platforms. In the coming months, you will be able to create your own account and use these generation features after the first demo.First demo coming this yearthen all the features described above will gradually become available throughout 2025. *You may not have access to some personal features.
