I second the motion with @sanchay...! Keep it going-Great Series! Thank you for including All the errors and fixes and your mental process using ChatGPT to assist with the bugs finding...
7 questions/suggestions: - what if we want to try out using already available videos of finite number (aspect ratio and landscape/portrait can be different from the desired final one) or available images (aspect ratio and landscape/portrait can be different from the desired final one) of finite number. The number of videos may be less or greater than the number of narration verses? - Like there should be some way to match the number of already available images or videos and put the available narrations on them to match the length of the end video. Have you given that a thought? A suggestion to make it even better. - Probably if the above 2 are not possible, some form of tool using OpenCV or an AI tool can be used to animate the images generated and the animation can be used in the video. - An option can be integrated to use LeonardoAI instead of Dalle. - As input parameter to the script photoreal, anime and other models can be given when Leonardo AI is used. - There can be a method at various stages to pseudo exit and first check the script generated (so as to edit it if required), then pseudo exit at the images generated so that they can be replaced if they look lame and then the script can continue onto the final video. - Probably there can be a method to add in background music from a file at 15% volume or so.
Thankyou so much. There's heaps in here for me to learn and I really appreciate you organising and explaining it how you have. Mark my words I will most definitely be buying you a coffee bro cheers ;-)
Hello bro I hope you are fine can I ask you a question about the project and I hope you answer me soon so the question is how can I edit the project to make it makes videos from the idea by likes it with copilot thanks
That was fun. I'm surprised you can get to sleep without hearing that narration over and over again mind you! If you're looking for something else to experiment with I've been playing around with the new 'Mixtral' model using the AnyScale API (it follows the format of the OpenAI one so is a drop-in replacement and you get 1million tokens for free). Mixtral is... "interesting". Bit like gpt3.5, but tends to follow instructions much more closely. Right up until it goes crazy and just spits out what looks like someone's job interview notes mixed with some FORTRAN. But hey!...
That sounds weird but interesting, haha... I'm more interested in the local Mistral 7B. It seems to run easily on my machine. Haven't really put it to the test, though