Introducing Sophie for Minimax H3.

Based on the same training data that was used for Sophie in Wan 2.1 and 2.2, this updated Sophie model includes a voice and all the stage direction possibilities that H3 can offer.

Important: This model is 100% fictitious and is not based on images of any real person. She originated with images generated from still image checkpoints and has evolved from there using only AI rendering. No images (neither actual nor AI-generated) of any real person was ever involved in the making of this model.

Recommended settings:

• keyword: sophie

• strength 0.7 to 0.8 most of the time

• sampler er_sde, scheduler sgm_uniform

• works with both turbo and standard H3

The strength of 0.7 or so helps keep the face from looking too overcooked, but it does cut back on some of her other qualities that you can emphasize in prompts to make up the difference.

• <d>[English with a clear British accent] Hi, my name is Sophie. I'm glad to meet you!</d>

• a sophie woman with short straight blonde hair

• if you want to emphasize or alter her body shape at all, do that with prompts as well

In my head, Sophie is English and speaks with a standard educated English accent like you'd see from a news broadcaster on BBC. Sometimes H3 gives her an American accent instead, but you can reinforce her accent with the dialogue prompt format shown above.

Sophie was trained with clothed and nude samples. As a result, if you talk about body parts, Minimax H3 will attempt to show you those body parts even if that doesn't make any sense. So be careful what you prompt for.

Be nice to Sophie and she will treat you well!