Tue Jul 11, 2023 12:01 am
You're most likely referring to my GPT and Resolve proof of concepts videos on YouTube. There is a video that shows a virtual assistant in Resolve, another one where we generate images in Resolve's timeline and a third one where we describe what a Fusion comp should do and then leverage GPT 3.5-Turbo to build the comp.
There is a fourth video coming out on July 11, but I won't say what it's about yet.
Several people have asked, but at the moment I don't have the time to make a how-to video.
On the OpenAI side I don't think you need a "Plus" GPT account. It's been a while since I've set this up and don't remember the details. I think all you need to do is sign up for an account, create the API key(s) and enter your payment information. The API pricing is per-use. There is no subscription.
To interact with OpenAI LLMs you can use their Python or Node.js API. You can even go low level and just use http GET/POST requests.
On the Resolve side the API is built in and free. Generally, it helps if you're using Resolve Studio. The functionality is the same except that Resolve can't take external API calls. Resolve Studio can.
As far as the interaction goes, OpenAI API is all about sending and receiving strings, except DALLE-2 and the audio transcription service. Whether these strings represent a natural language or something else like computer code is up to you.
You can do two things. You can either fetch some strings from Resolve and send them to OpenAI LLM or DALLE-2, receive a response and then send that response to Resolve.
Or like in the case of the Fusion comp builder, you can formulate some string data outside of Resolve, send it to OpenAI, then use the response to do something in Resolve.
www.metafide.com - DaVinci Resolve™ Apps
