TL:DR Input articles to generate a script, generate voiceovers based on the output from the script, generate background music from the output, input videos to download/clip/randomizes/compile, then put it all together to make one video all at pushes of a button no extra work needed
I built this specifically for recap or news niches so you cant just make whatever you want for whatever niche but eventually I want to get there this is just the first version.
So good niches this will work on is celeb news, movie, sports, talking about streamers, rock n roll, animal, spiritual, ai stories, history, tv shows, and anything where the contet being watched doesnt need to follow what the words are being said.
After creating a project, you load your articles from wherever it can be multiple links or one and then press generate. This makes an api call to chat gpt to create a lengthy 30-minute script, chucks it out for voiceovers, title, description, tags, and music generation. This gets saved in your folders where you specified at the start of the project don't worry you don't waste any api calls because you can't move forward without saving a project spot.
We always have records of what is being generated just incase you want to make changes at any step. In the audio step you can generate your voice overs based off the scripting output. You can choose the voice you want as they come from 11 labs. Then you can generate your background music based off the music prompt generated. It gets created from the vibe that the article is that you submitted so its tailored to the video. We use suno AI to create the music and because suno doesn't have an api we use a wrapper from kie ai.
The background music and the voice over clips and the compiled voice over clip gets generated and saved in your folders for review and later use for the putting it all together.
The last tab for inputting information is the video tab its here you input your youtube/local videos it will download them up to 4k quality output. The proces works like this download, clip the vide less than 6 second intervals, randomizes them and then recompiles them to make one video and then upscales it 4k and increases the bitrate based off the resolution. This is why there is a very specific use case right now and doesn't just make any kind of videos but by doing this we are youtube compliant and will not get triggered for copyright.
When you have all the necessary requirements to make a proper video, voiceover, bg music, and recompiled video we put it all together to make a completely new video ready for publishing. The only thing a user really has to do is create the thumbnail. A complete 20-30 minute video is created in less than 10 minutes well depending on which resolution you choose. Things do take longer at 4k