HOME | DD

Description
I have been working for a loooonnng time trying to train a model to generate some wrestling hold and I'm finally getting somewhere. The photo was 100% generated from a text prompt, this is pure text to image. Nothing else but text was used, no control net, instant id or any other tool was used. The image wasn't even refined, this is the raw output !

The model is far from perfect: it only handles sleeperhold (I had to start with a hold), about 1 in 4 images are correct so there is still some cherry picking to do. But nonetheless, it finally works ! I can't wait to improve the model and train on other holds (I have a whole dataset ready !)

Here is the gallery related to this model to see other examples: www.deviantart.com/unexistor/g…

Description
I have been working for a loooonnng time trying to train a model to generate some wrestling hold and I'm finally getting somewhere. The photo was 100% generated from a text prompt, this is pure text to image. Nothing else but text was used, no control net, instant id or any other tool was used. The image wasn't even refined, this is the raw output !

The model is far from perfect: it only handles sleeperhold (I had to start with a hold), about 1 in 4 images are correct so there is still some cherry picking to do. But nonetheless, it finally works ! I can't wait to improve the model and train on other holds (I have a whole dataset ready !)

Here is the gallery related to this model to see other examples: www.deviantart.com/unexistor/g…