The inspiration for this project came from me simply scrolling through YouTube Short’s. Some of the shorts I ran across were small snippets of Tv series. As I was watching them closely I realized that making these short videos actually requires a lot more than what the average viewer may notice. From that moment I decided to try and make a tool that could significantly help with this process!
HILYT, the name of the project, is in essence a tool that anyone can use to completely automate the YouTube shorts process. To properly demonstrate all its capabilities I will run through a scenario. Let’s say you have a playlist of YouTube Videos, and you want to make shorts out of all of the videos. What HILYT does is it starts from episode 1, downloads the episode, transcribes the video, picks the best moments (using Google Gemini) and forms them into shorts that are cropped to your liking, have your own background music, video filters, word by word captions, jump cuts and more. It takes the clips made ( the number of clips made can be determined by you ) and it uploaded these clips to your YouTube channel and schedules its release times (which you can also choose.) Then it automatically moves to the next episode and continues this process until the entire playlist has been used to make shorts.
This project was built using a multitude of different tools and applications. Here is a list of just some of the tools and what they were used for: The overall project was mainly built using Python. OpenAi’s Whisper model was used transcribe the text. Google Gemini was used to pick the best moments, and Demcus was used in the case that the audio needs cleanup. Building this project required the culmination of all these tools and more into a system which worked.
Throughout the course of building this project there were many complications that I ran into, so I will just mention 2 of the more prominent issues. One of the big ones that I had was the YouTube guidelines. To test the capabilities of my project I decided to use an animated TV series playlist I found on YouTube. My goal was for the shorts to all have enough transformations for it to fall under fair use while being enjoyable. The problem was that the series I was using had background songs/audio which caused the shorts to be blocked whenever it was uploaded, despite the fact that I put filters and cropped the video. To solve this the solution I used was a tool called Demcus. With this tool, the background songs from the original video was removed and all that was left was the dialogue. Then from there I was able to add my own copyright free music as the background. Another issue I ran into was with the scheduling process. Let’s say that you wanted to make 12 clips from the first video in the playlist. The problem is that YouTube has a limit to the amount of videos you can upload in a day. As a result uploading 12 clips may not work. The fix to this was to have the user choose how many clips are uploaded at once. Let’s say the user decided they want 4 clips uploaded at 4 different times in the day even through the made 12 clips. What HILYT does is it schedules the 4 clips for day 1 then waits for YouTube’s quota to reset. Then it schedules the next 4 for day 2. It waits again. Then it schedules the last 4 for day 3. After that it moves on to episode 2.
There are many accomplishments that I am proud of regarding this project. I would say my biggest pride in this project is creating the world’s first system which can sequentially crawl through each episode of a playlist make shorts out of them and upload it to your channel. What sets HILYT apart is its ability to automate the clip production process, and the sheer amount of customizability that went into it. I believe this system will be a huge help to anyone in the content creation space.
While creating this project, I would say that the number 1 thing I learned was patience and the ability to always step back and look at the bigger picture. Because of the multifaceted nature of this project, it is very easy to end up going down rabbit holes which often leads to dead ends. The ability to step back, look at the big picture, make compromises, and be patient was crucial in me being able to get as far as i did in this project.
For now, I will focus on adding some finishing touches to the project. If I move forward in this Hackathon competition then my plan will be to create the front end and implement this project fully into Adobe Express.
Log in or sign up for Devpost to join the conversation.