Features

Model selection is a core feature of DecentAI that puts the power of choice in your hands. At any point during your interaction, you can tap the model selection button at the top of the screen to access a wide range of AI models and model mixes.

When you tap the button, a bottom sheet appears, presenting you with two main options: model mixes and individual models. Model mixes are specially curated groups of AI models, assembled based on their performance characteristics or quality. For instance, the "Small and Fast" mix includes models with fewer parameters or those hosted by inference providers optimized for speed. DecentAI's intelligent routing system automatically selects the most appropriate model from the mix for your specific query.

If you prefer more control, you can choose to use an individual model instead of a mix. This allows you to consistently interact with a specific AI model of your choice.

To help you make informed decisions, each model and mix comes with an information icon. Tapping this icon provides more details about the model or mix, including its capabilities, strengths, and potential use cases.

The model selection feature is designed to let you explore different AI models and find the ones that work best for your needs. Whether you're looking for speed, accuracy, or specialized capabilities, you have the flexibility to experiment and discover your preferred AI interaction experience.

Voice Chat lets you talk with AI models using your voice instead of typing. It's straightforward - you speak, the app converts your words to text, sends them to the AI, and then turns the AI's response into speech.To use it, tap "Voice Chat" on the home screen or look for the headphones icon in a conversation. Then just hit record and start talking. The app uses a mix of open-source and commercial models to handle the speech-to-text and text-to-speech parts. It's designed to be quick, so you shouldn't notice much delay between your questions and the AI's answers.This feature can be handy if you're multitasking or just don't feel like typing. It might also be useful for people who have trouble with text input.

DecentAI's "Bring Your Own Model" feature offers an advanced level of customization and control over your AI interactions. This functionality allows you to connect directly to your personal AI models, whether they're running on a local device, your personal computer, or deployed to a private cloud service.

To use this feature, start by tapping "Add your own model" in the model selector. You'll need to provide an OpenAI-compatible endpoint (typically a v1/chat/completions URL) and an authentication token if required. Once set up, DecentAI will establish a direct connection to your specified server, enabling you to chat with your preferred language model.

This feature is particularly useful for users who have:

• Fine-tuned models for specific tasks or domains

• Models running on local hardware for enhanced privacy

• Custom deployments on private cloud infrastructure

By allowing you to integrate your own models, DecentAI provides unparalleled flexibility and sovereignty over your AI experiences. You can leverage the app's user-friendly interface while maintaining complete control over the underlying AI model and its hosting environment.

Remember that when using your own model, you're responsible for the model's performance, availability, and any associated costs of running the infrastructure. DecentAI simply facilitates the connection between you and your model.

Image Generation mode in DecentAI provides a dedicated space for creating high-resolution images using AI. This feature allows users to harness the power of various image generation models, including cutting-edge options like Stable Diffusion 3 and DALL-E 3, among others.

To use Image Generation mode, simply enter your desired prompt in the text field at the bottom of the screen. You can then select from a range of available models to generate your image. The app presents the results in an infinite feed, allowing you to continuously explore different visual interpretations of your prompts.

When you find an image you like, you can long-press on it to bring up options for sharing or saving the image locally on your device. This makes it easy to use your AI-generated images in other applications or share them with friends and colleagues.

For those who want to review their creative history, the gallery view offers a comprehensive display of all generated images in one place. This feature allows you to easily browse through your past creations and revisit your favorite generated images.

As you use DecentAI and provide feedback, you can earn points as a reward. As a start, this points system recognizes active participation within the DecentAI and open source AI ecosystems. We will share more info as we continue to evolve and expand the system.

We believe that decentralization goes beyond just open model weights. We want to provide you with the option to execute certain models on decentralized GPU networks like Heurist and Kuzco. By leveraging these networks, we take a significant step towards open and disintermediated access to powerful AI models, ensuring that you have the freedom to utilize the full potential of AI without relying on centralized systems.

We are exploring Apple’s planned option to deploy tuned private models for iOS access, as well as integrations with providers such as Hugging Face, allowing you the option to connect to any model available on the platform. This helps you stay at the forefront of AI innovation and run inference on the latest models as they’re released, and choose models designed to handle specific specialized tasks.

We are committed to integrating open source models that allow you to interact with AI using various modalities, including audio, images, and video. On a rolling ongoing basis, we plan to integrate more generative models and enriched app UX, giving you a wider range of options for creating images, music, and video. Our goal is to make these experiences available in a fully open source manner. 

Everyone has unique personal requirements and customization needs when it comes to AI. To meet these needs, we plan to provide greater control over parameters, prompts, settings, tools and data usage. With DecentAI, you should have the ability to create your perfect open assistant aligned with your preferences and goals.

DecentAI is dedicated to furthering the open source development of AI models. To support this goal, we plan to expand our dashboards to include additional model and provider metrics. By sharing this data transparently, we aim to foster collaboration and drive the advancement of open source AI technologies. Importantly, no personal chat content will ever be accessed by these dashboards – because we don’t have that data. We track only anonymous events data such as model and provider selection.