This is a very complex process. I get it. I want to be constructive. It seems like very basic film terminology isn’t recognized by Sora. I’ve been a film editor for 40+ years, and it’s not that I’m set in my ways, it’s that there are simple, precise ways to describe shots, that objectively help. When I say I want the reverse angle of something, for example, Sora seems not to get it. When I say that I want the character to deliver lines facing “screen left,” it’s the same problem. Sora might want to sit down with a film editor, script supervisor, director, or all of the above to incorporate some basic terms.
3 Likes
Hi, welcome to the community!
And I upvote this, or whatever, and think that the audio responses should similarly be trained on vocabulary from the voice direction industry. I think it would be the most helpful if common visual / voice directions were hardcoded as a part of the API.
There are lots of times when you may want to give voice direction, like the strength of an accent. In Audio Responses, it’s all still non-deterministic. So, if you ask for a “strong French accent” you might literally get anything with each generation—not great for narration.
We need to be able to stabilize these types of situations according to the vocabulary in the industry (film / recording).
1 Like