So today I was looking for a way to create a video with AI and found a number of soutions:
[invideo.io] (https://invideo.io/) [aistudios.com/] (https://aistudios.com/) [hubspot.com] (https://hubspot.com) [synthesia.ai/] (https://synthesia.ai/)
My aim was to produce a simple introductory video for a training course that I’m developing. Over the xmas holidays, I made use of [suno.com] (https://suno.com/) to create an AI song about my daughter and her friends on a ski holiday. I then pain stakingly put a video together of our trip and edited the video as best I could to the song lyics. This got me thinking that there must be a better way to create a video. So first step was to hunt around the net to find out what tools are available.
TLDR: Invideo.io seemed to provide the best results, although it’s a paid service. If you don’t pay, then videos will contain watermarks and use some standard stock imaagery.
AI Studio, despite it’s high rating on google search seemed to struggle somewhat. The output video only contained 2 static images and the spoken audio was not great, mispronouncing words every now and then.
Hubspot also looked good, but it’s a free service, but you have limited themes. It’s mostly setup to help promote your company or business so it felt that the output wouldn’t really match with the topic of my script.
Synthesia.ai looked good, but after inputing my prompt and email address, I received an email saying my prompt was in-appropiate and the video would not be generated.
This then prompted me to see if I could build my own video generator. After all I have access to a couple A6000 GPUs which have been somewhat idle. So I set about trying to build something using python and streamlit. My workflow was to use o1 to generate a high level requirements spec. Then this was fed into claude to generate the code. I asked claude to ask me any questions in case of any ambiguity.