Sora Image to Video issues (lack of control and motion)

ChatGPT
,
1

I’ve uploaded hundreds of images to test out image 2 video. And control over the output is basically nonexistent. Meaning the output is 99% static or with extremely strange camera movements. People are mostly static lacking even basic subtle movements. Basic camera movements like zoom in zoom out, pan up pan down, dolly in etc aren’t responsive to commands. You occasionally get some camera movements but it’s erratic and unpredictable. And when you do finally get movement most of the time it transitions from your image to a subpar sora recreation of the scene which stylistically looks very different.

I’ve uploaded without prompting, adding prompts, editing storyboards, adding little notes on multiple storyboards, removing or editing the sora created storyboard that is auto-generated when you upload an image. And there’s little to no control and generally speaking very little movement or strange movement or just hallucinations in the output.

Sora seems to handle text to video well if you follow the sora prompting guides. But image 2 video is severely lacking. Can’t even get a simple zoom in, zoom out, dolly in etc. I just want my images to have a little life and some very simple camera moves but it seems rather impossible at this stage.

Please share some thoughts if you’ve gotten it to work consistently. I’ve tried many different types of images with and without people different artistic styles of images etc. And the results are abysmal and non reactive to prompts with image 2 video.

Please help. Thanks.

4 Likes
4

I have the Pro version and am having the same issue. In one of my results the photo was exactly the same as a stagnant shot. The other was the exact same shot but with a strange zoom in. This function of Sora doesn’t seem to be working at all.

As far as the text to video function, I am trying to use Sora for a short film project for a nonprofit client and so far have had very disappointing results. It seems to work great for flying creatures or woolly mammoths romping in the snow, but so far after spending hours and many “credits,” I have not been able to get any shots I can likely use in this project. Lots of strange artifacting and not following prompts.

Sadly, Sora feels like it’s still in R & D mode and not ready for the general paying public.

3 Likes
5

Exactly my experience, although I’ve had some satisfying results.
My latest attempt was what I thought would be relatively simple and yet cool: “an old man is painting a woman who comes alive and steps out of the canvas”, well guess what? All I got was an old man painting a woman with camera movement. So I got rid of the old man: “a woman in a painting suddenly comes alive and steps out of the canvas”. I even used the AI generate text to enhance it. Result? A painting on a wall of a woman with weird camera movements, otherwise completely lifeless.

3 Likes