A problem I've always found with image prompting is the extreme reluctance of most models to actually put someone wading properly deep in a substance. Ask for "wading waist deep" and mostly it barely comes up to their shins, I found swapping "wading" for "swimming" helped a bit - at least to thigh depth, but going any further remained a major challenge.

Promptus recently added ByteDance's SeeDream 5.0 as a model, which in general produces superb image results, so I sat down at the Mission Control PC in the eastern tower, with it and Gemini, and investigated how to get someone shown at different depths in custard. I started with with thigh deep, and then lowering them down (or raising the custard level) to see them with crotch submerged, then waist deep, abdomen submerged, chest deep, only shoulders dry, and finally neck deep. I told Gemini exacly what I was after and it was very helpful in explaining use of terms like "conceal" to specify which areas of the person are out of view - obvious with hindsight but not something I'd thought of previously. It also confirmed what I'd long suspected, if you specify the footwear the generators will lower the level so the shoes show - best to just not mention anything that isn't meant to be visible.

Another insight it gave is that every sentence has to be thought of as a specific instruction to the generator, as opposed to the way a human will interpret an entire paragraph overall. Again, not something I'd previously considered.

Attached are some of the best of the results of the tests I ran, and below are all the prompts used, in some cases several times or with minor changes (which I failed to document, oops!) to achieve these results.

I used https://promptus.ai with the model set to "(ByteDance) SeeDream 5.0 Text to Image". Note it tends to default to SeeDream 4.5 which isn't anything like as good at faces, so check (esp if browessr has restarted) that it's still on 5.0.

And given if you ask for "woman" you invariably get white westerners, I deliberately chose a differen ethnicity for the test person to bring some diversity. The surfing swimsuit was also to test how it'd be with a relatively sexy costume, I got the occasional moderation flag but resubmitting the same prompt usually went through OK on second attempt.

[examples of prompts]

A focused medium shot of a Japanese woman standing in a large vat filled with thick, viscous yellow custard. She is wearing a grey and turquoise long-sleeve surfing one-piece swimsuit. The deep, heavy custard is entirely opaque and and comes up to her thighs, completely concealing her lower legs. High-key studio lighting highlights the messy, viscous textures of the food. She is smiling and looks comfortable, clearly happy in the mess.

A focused medium shot of a Japanese woman standing in a large vat filled with thick, viscous yellow custard. She is wearing a grey and turquoise long-sleeve surfing one-piece swimsuit. The deep, heavy, waist-deep custard is entirely opaque and completely conceals her legs. High-key studio lighting highlights the messy, viscous textures of the food. She is smiling and looks comfortable, clearly happy in the mess.

A focused medium shot of a Japanese woman standing in a large vat filled with thick, viscous yellow custard. She is wearing a grey and turquoise long-sleeve surfing one-piece swimsuit. The deep, heavy, waist-deep custard is entirely opaque and completely conceals her lower torso and legs. High-key studio lighting highlights the messy, viscous textures of the food. She is smiling and looks comfortable, clearly happy in the mess.

A focused medium shot of a Japanese woman standing in a large vat filled with thick, viscous yellow custard. She is wearing a grey and turquoise long-sleeve surfing one-piece swimsuit. The deep, heavy, chest-deep custard is entirely opaque and completely conceals her lower body. High-key studio lighting highlights the messy, viscous textures of the food. She is smiling and looks comfortable, clearly happy in the mess.

A focused medium shot of a Japanese woman standing in a large vat filled with thick, viscous yellow custard. She is wearing a grey and turquoise long-sleeve surfing one-piece swimsuit. The deep, heavy, chest-deep custard is entirely opaque and completely conceals her lower body. Only her shoulders, upper arms, and head are visible above the messy, swirling surface. Her lower body and legs are completely invisible, submerged deep below the surface of the custard. High-key studio lighting highlights the messy, viscous textures of the food. She is smiling and looks comfortable, clearly happy in the mess.

A focused medium shot of a Japanese woman standing in a large deep vat filled with thick, viscous yellow custard. She is wearing a grey and turquoise long-sleeve surfing one-piece swimsuit. The deep, heavy, shoulder-deep custard is entirely opaque and completely conceals most of her body. Only the tops of her shoulders, neck, and head are visible above the messy, swirling surface. Her body and legs are completely invisible, submerged deep below the surface of the custard. High-key studio lighting highlights the messy, viscous textures of the food. She is smiling and looks comfortable, clearly happy in the mess.