Unfortunately, visual input is just simply too slow. It creates a noticeable delay between voice clips in the commentary, and the commentary is always behind the action. This, in combination with the model's tendency to say things like, "As you can see in the image," this is not a viable feature for the time being.
That said, the program has been set to always use the latest GPT-4 Turbo model, with a variable self.model in the TextGenerator class for easy updating down the road. The user can no longer choose a different model. This will help keep the output consistent, and is a reasonable change given OpenAI's recent price drops overall. Incidentally, this also solves the issue of the GPT model not switching after selected it in the settings, since the user can no longer do this.
The library requirements were also updated for this PR, as a couple libraries are no longer needed. There was also a security concern with the Pillow library, so it and all other libraries have been updated to their most recent releases.
Fixes #60
Fixes #41
Type of change
[x] Bug fix (non-breaking change which fixes an issue)
[ ] New feature (non-breaking change which adds functionality)
[ ] Breaking change (fix or feature that would cause existing functionality to not work as expected)
[ ] This change requires a documentation update
Testing
Commentary was run on a known good replay for commentary generation, with several approaches to try to make visual prompting work before deciding to remove it. The commentary generation works as expected after all the changes.
Description
Unfortunately, visual input is just simply too slow. It creates a noticeable delay between voice clips in the commentary, and the commentary is always behind the action. This, in combination with the model's tendency to say things like, "As you can see in the image," this is not a viable feature for the time being.
That said, the program has been set to always use the latest GPT-4 Turbo model, with a variable
self.modelin theTextGeneratorclass for easy updating down the road. The user can no longer choose a different model. This will help keep the output consistent, and is a reasonable change given OpenAI's recent price drops overall. Incidentally, this also solves the issue of the GPT model not switching after selected it in the settings, since the user can no longer do this.The library requirements were also updated for this PR, as a couple libraries are no longer needed. There was also a security concern with the Pillow library, so it and all other libraries have been updated to their most recent releases.
Fixes #60 Fixes #41
Type of change
Testing
Commentary was run on a known good replay for commentary generation, with several approaches to try to make visual prompting work before deciding to remove it. The commentary generation works as expected after all the changes.