A Journey Through 2024
2024 saw a lot of changes for DeepFakeAI, as our project continues to mature and develop new features. We’ve passed our 1 year anniversary, generated thousands of new videos, and accumulated a whole new cast of characters to use. Follow our 2024 wrap-up as we ring in the new year!
Where We Started
At the end of 2023 we were already on our way to the moon, boasting over 3,500 Telegram members in our main group, consistently putting out our in-house trained characters throughout our first year up and running, and in under 200 days generating over 30,000 videos! Our character polls had begun for token holders, and we had been mentioned on Joe Rogan, Snopes and in several written articles.
New Year, New Features!
As January began, several new characters joined the DeepFakeAI family as per our stakeholder polls. Meanwhile, our team was hard at work behind the scenes bringing new features, and in February one of those new features were released, our HD Lipsync Model! It vastly improved the quality of deepfakes made with our platform, and has since been further developed to fine tune our generated videos.
Dashboard v2
We had been working with a model where the “Dapp” and “App” were two different things, which was confusing and inefficient. To make the interface better, we combined the Dapp and Web App into one unified Dashboard interface in March, which in the coming months would have an expanding list of tabs containing all of DeepFakeAI’s features in one neat package.
A New Paper for the New Year
May saw the publishing of a new Whitepaper for DeepFakeAI, with a revived focus on the technology featured on the fakeai.io website and in the FakeAI_Bot. It replaced the previous whitepaper, which focused mainly on the use-cases for the service and the integration with cryptocurrency.
Do It Yourself – The Big Release
In May, we released our most revolutionary feature to date – DIY Character models! This feature allows users to train their own character models, vastly increasing user creativity options, and was a huge hit from the moment of its initial release to coin holders. It’s since been expanded in a number of ways, as you will read below.
Do It Yourself, Together!
Within a week of the DIY feature’s release, the Explore tab was introduced on the Dashboard, which allows users to share their trained character models with each other. This change upgraded the whole experience of using the service and allowed users to expand the character pool themselves, forming a whole ecosystem of creation, sharing and clip creation.
A Marketplace of Character Models
In June, the Rating System was added to the explore tab, which allows users to vote characters up or down based on how good the character model is to use. This allows the best character models to float to the top, and users to get an idea of which characters they may want to use among the many models available.
Fine Tuning for Everybody
July saw the release of the fine-tuning feature, which allows users to further fine-tune their character models after their initial training. The further training allows the models to become higher quality without the inconvenience of creating a whole new model from scratch.
A Huge Development
After the release of Fine-tuning, the team got down to work, developing a huge improvement to the way that character models are created. Previously, the characters were created using traditional training models, which took a significant amount of time and resources. While the release of custom characters was a great accomplishment, just around the corner was a giant step in character training development.
The Zero-Shot Model
The Zero-Shot Model had been teased for quite a while, but finally saw its debut in November. This allows new characters to be trained in minutes, and with the decrease in time and resources required to train new models, the DIY character feature was released to all users for free! Now any user who wants to use the service can train their own character models in no time at all, in high quality, all based in the cloud!
Bots for Everyone!
Another new feature released in November was the revamped Bots system, which allows all users to create their own bots and give them all of the features of the old FakeAI_Bot. This gives users complete control over how their bots are used, and means a greater level of customization for each server’s assigned clip-generator.
Something for the Readers
Late in the month, Captions were added as a feature, allowing better readability on mobile and accessibility for the hearing impaired. Additionally, in December, the DeepFakeAI Gitbook was relaunched, with new tutorials and all the information one could need about using the platform. This created a hub to point to when any user has any questions, and improved the service’s ease of use.
And that’s a wrap!
And with those releases, the year is coming to a close, and we want to thank all of our users for the continued usage and feedback. We have so much more to release, which we are excited to announce going into 2025. Stick around to find out what our next steps will look like, and keep on creating!
Leave a Reply