AI-Flow is a platform for creating custom AI tools through a simple drag and drop interface.
Run AI workflows in the app or integrate them into your projects.
AI-FLOW is an open-source platform that enables you to connect multiple AI models effortlessly using a simple drag-and-drop interface. Create custom AI tools in minutes, without any coding knowledge. Try it now for free and unleash your creativity!
In May 2023, I embarked on a project with a simple goal: to enable anyone to connect various AI models together without needing any coding knowledge. It all started with my desire to generate illustrated stories using AI, but I quickly grew tired of constantly switching between ChatGPT and MidJourney.
As I delved deeper, I realized the potential for more applications. For instance, summarizing webpages and YouTube videos should be straightforwardājust give the content to GPT, ask for a summary, and maybe refine the prompt. Why pay for something that can be done in minutes? Plus, I'm already paying for OpenAI...
That's why I created AI-FLOW. It's an open-source tool that allows you to seamlessly integrate different AI models. The core version is available as a Windows executable and is perfect for basic use cases.
For those looking to unlock the full potential of AI-FLOW, some setup is required. However, if you prefer a hassle-free experience with automatic updates, access to all models, and unique features, our hosted version is perfect for you. It uses a flexible credits system, and if you set up your own API keys, you can significantly lower your costs.
With AI-FLOW, you can now enjoy the power of connecting AI without the complexity or cost.
@dahnm20 Congratulations on the launch! I've had this struggle when building the product I'm launching soon, actually! Ended up having to write a lot of code to automate running a text-to-image interference from my own model, with a prompt expansion by LLM, which then is fed to faceswap model, then GFPgan facefix, then upscale lol. This tool could've saved me so much time!
@ai_guitarist Thank you!
Yes, that's exactly it! Nowadays, we often need to chain a few models together to achieve good results. I'm currently working on making it integrate easily with other tools, which might help people in building products.
I just tried out this product, and I love it! First of all, thank you for allowing a short trial without the need to sign-up. I am looking forward to using your product more. What I don't understand is the pricing model, how do your charge your customers?
Also, just as additional feedback, please allow for the users to add their own fine-tuned models. This is a game changer, and it isn't a hard feature to implement for you. I have seen many drag and drop tools that tend to be similar (like voiceflow), but they don't allow you to add your own fine-tuned models.
While people could technically use Replicate for their fine-tuned models, I feel like adding it as a special feature for gpt can be of great help.
@mariusmanolachi
Hi Marius,
Thanks!
Regarding the pricing model, at some point, it will be mandatory to log in to use the hosted version of the app. When logged in, you will have access to a credit system. With the credits, you'll have access to every tool.
However, we understand that many users have their own API keys and do not want to pay for a duplicated service. Therefore, you have the option to add your own API key and use the related nodes at a lower cost.
For example:
If you generate an image with Stable Diffusion 3 without your key, you'll spend 3 credits. If you have your key set up, you'll spend only 0.1 creditsāa small fee for resource usage.
@mariusmanolachi Just to be sure, do you want to be able to use a fine-tuned GPT that you've created with the OpenAI API? Or do you want to add a fine-tuned model that you're hosting elsewhere and use it in the same way as the GPT node?
Great job, lots of potential. A must-support app, definitely. I would suggest to refine the web extractor node: right now it arbitrarily extract content, often text from cookies alert, hence not much useful. It would be great to have control over selecting page elements. If you want an example, I was trying to feed ChatGPT 4o the first 3 results of a Google news search and yes: I've got Google cookies alert text in French (where your server is hosted maybe?) as a result. I was hoping as a work around to be able to tell ChatGPT node to selected first three paragraphs or divs with a certain class, but the web extractor node didn't seem to have stored the whole page content.
Still, I'm definitely going to use this, and I'm supporting it.
@fabiosalvadori
Thank you so much for the detailed feedback!
I think it's a great idea to rework the node to add options for personalizing and better controlling the extraction. I hadn't thought of that before, but I will definitely work on it ! Thanks !
AI-Flow
notclass
AI VisionBoard
AI-Flow
notclass
notclass
AI-Flow
AI-Flow
notclass
AI-Flow
Wallafan
AI-Flow