This LoRA will help you construct multi-frame story sequences with breast expansion. This is a tricky concept to train in a LoRA so I'm sure this first effort is far from perfect. However, training and tweaking in Flux with my 3060 GPU is painfully slow so I'm calling this "good enough".The keyword is "besequence", which I've used in the filename for the LoRA safetensors file to help you remember. I'll include some sample prompts below as well. The easiest way to duplicate my example images is to use ComfyUI, which is what I used to create them. The images have m workspace embedded, so all you have to do is download the full source image and drag it onto your ComfyUI workspace. You'll see the prompt, the seed... everything. If you want to use that workflow yourself, you may need to download some extra custom nodes, but if you just want to inspect an image for the creation parameters, you won't have to install anything just to look and it's really easy that way.The LoRA also works to some extent as a general purpose transformation LoRA by tweaking the prompts and optionally adding other LoRAs to your pipeline. I've seen - on purpose and by accident - muscle gain, weight gain, and other changes in my tests. Flux is pretty good at separating desciptions for each panel in a multi-part series, so you can use that ability to tweak a specific panel in one direction or another.As I said, the LoRA (or Flux) isn't perfect, so you can expect to have to make several examples with different seeds to get a series where all the parts are just the way you want.<h3 id="example-1:-illustration-series-of-a-woman-drinking-a-potion-j2ku3x7uy">Example 1: Illustration series of a woman drinking a potion</h3>Prompt:<pre><code>A comic book story told in 3 besequence panels, flat color comic book illustrations of the same young woman standing in her kitchen wearing a long pink t-shirt and jeans. Her body is thin and athletic.</code></pre><pre><code>First panel: She has a flat chest with very small A-cup breasts and her shirt is very loose on her body. She is drinking a glass of bubbly green liquid.</code></pre><pre><code>Second panel: She has medium-sized C-cup breasts and her shirt is a little tight. An empty glass is sitting on the counter.</code></pre><pre><code>Third panel: She is extremely busty stretching out her very tight shirt. Her body is very thin and athletic. Her hands are touching her chest. She is happy and smiling.</code></pre>In this example, I'm calling out each panel with specific instructions. Flux understands a variety of descriptors like:<ul><li>Panel 1, Panel 2, Panel 3 - has a chance of Flux adding number labels to each panel</li><li>First panel, Second panel, ...</li><li>Left panel, Middle panel, Right panel</li></ul>Depending on how many panels you're using, you can try one or the other. If you don't want to describe each individually, see the next example below.<h3 id="example-2:-woman-in-red-blouse-h2mwrlp7l">Example 2: Woman in red blouse</h3>Prompt:<pre><code>5 besequence panels, photos of the same beautiful woman in an office wearing a red silk blouse with a grey wool pencil skirt. In the first panel she has flat a-cup breasts and a thin body with a neutral expression. From left to right her breasts grow bigger and bigger. In the last panel her blouse is unbuttoned and she has enormous breasts with cleavage and a happy smiling face.</code></pre>Here, my description is more vague and leaves more up to Flux's imagination. I described the beginning and end and hoped for the best. I had to generate more samples of this kind of prompt before I got one I liked that showed a good clear progression. This format is nice for images with more than 3 panels where writing all the descriptions is more inconvenient.<h3 id="example-3:-supergirl-and-the-growth-ray-itetkpzin">Example 3: Supergirl and the growth ray</h3>Prompt:<pre><code>3 besequence panels, photos of the beautiful woman wearing a supergirl costume with long wavy blonde hair standing in a dark city alley. She is thin, fit, and has a muscular athletic body.</code></pre><pre><code>First panel: She has a flat chest with very small A-cup breasts. Her body is outlined in glowing green from a green ray of light hitting her from above.</code></pre><pre><code>Second panel: She has medium-sized C-cup breasts and her costume is a little tight. She is looking down at herself with surprise.</code></pre><pre><code>Third panel: She has very huge enormous breasts stretching out her costume with deep cleavage. Her body is thin athletic and fit.</code></pre>More of the same here. You can see the different panel descriptions are used to tell more of a story than the simple red blouse image.General advice:<ul><li>using the words "huge" or "enormous" or similar will sometimes turn into an "all over the whole body" effect, producing weight gain or a different body shape. You can usually correct for this by re-emphasizing the body shape you want in each panel description. Unless you WANT all over growth, in which case that should be easy to achieve.</li><li>Sometimes Flux gets the growth sequence out of order. Nothing much you can do here except try another seed.</li><li>If you get an image with a weird anatomical artifact like an extra arm or something, try rendering with an extra 5 steps and it usually fixes it.</li><li>You may need to also call out other specifics in each panel if Flux adds changes from one panel to another that you didn't want, like a different hair style for example. I didn't always bother in my examples, but if you want to aim for the perfect picture that's something you can do.</li></ul>In spite of everything I've said about Flux's prompt reading skills, the LoRA does help. I'm including a couple of comparison images in the LoRA samples at the top.I hope you enjoy using the model!

Files with SHA256: 13c6893183a764989e2bd19297bbb8af7bc200d23fdf01c5eda436014e22b7d6