I need a consistent(same) character but with different poses.
This is literally the lament of every dev trying to use AI to make game assets - inconsistency in details of face, hair, clothing.
The two closest approaches that I know of are:
1 - use the name of a well known public figure in your prompt. The more common the public photos of the person are, the better this works. (Note that this is why so many early ai generated images looked like Emma Watson- she had a LOT of images in the training sets of early stable diffusion). Downside is that everything will look like a known person which can be weird.
2 - use a source such as Daz to render a consistent character in the poses you need, then feed that into StableDiff via control net openpose and controlnet canny. The daz render does not need to be "good" visually, it just needs a clear body position, and face details. As a bonus you can often get quite consistent clothing details to come through, which is actually hard to do with pure prompt based renders.