<h3 id="v2.4">v2.4</h3><p>Thanks to <span class="MentionNode_mention__q78sK" data-type="mention" data-id="mention:567528" data-label="Foredev">@Foredev</span> for the tip, it funded a great deal of experimentation. This version finally produces results I like better than v1 but I am curious to hear from you all how it compares in your use.</p><p>Prompting specific features at full strength, such as facial expressions, can overpower the see-through effect. Reducing prompt weight, e.g. "(smile:0.7)", can produce results that successfully interpolate between both concepts.</p><h3 id="v2.3">v2.3</h3><p>I have tried without success to improve the model. This version lacks some of the older artifacting but is less consistent producing the effect. If I am going to spend all this buzz training bad models I might as well release one of them for you to try. Maybe you will find a use for it.</p><p></p><h3 id="v1-large">v1 - Large</h3><p>The same dataset as v1 - Normal on a lora that is twice as large. Results seem promising, but your feedback would be appreciated.</p><p></p><h3 id="v1-normal-(recommended)">v1 - Normal (Recommended)</h3><p>With an enhanced dataset the results are improved. There is still a strong style influence which biases the results towards an anime style and backgrounds exhibit overfitting. Effect works best when the background is a repeating pattern (like stone walls, concrete paths) or includes elements that extend across the subject (windows, edges etc). Although it was not tagged for this at all, adding "<em>see-through head</em>" may improve the transparency of the head when used in addition to "<em>see-through body</em>".</p><p>Adding tags like "<em>see-through clothes</em>" to the negative can get a more invisible girl style. Some issues may also be solved by adding "<em>tattoo</em>" to the negative.</p><p></p><h3 id="v0">v0</h3><p>Proof of concept: This is a proof of concept for training a concept on synthetic and modified imagery. The initial dataset was very small and the resulting model is overbaked but shows signs of concept comprehension.</p><p></p><h3 id="method-explainer">Method Explainer</h3><p>v0 and v1 were trained entirely on synthetic data created by generating a subject and a background independently and then compositing a semi-transparent subject over a background using image editing software.</p><p><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/3e814023-4bbc-4d62-a4cc-bee7f7878c33/width=525/3e814023-4bbc-4d62-a4cc-bee7f7878c33.jpeg" /></p><p>This approach has an obvious downside, that the two components are not well integrated. How then to create a full body subject oriented correctly to the viewing angle of the background? My solution was to turn to my favourite tool, the lineart controlnet.<br /><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/5fda840d-2708-471e-a10b-8dd298d0e934/width=525/5fda840d-2708-471e-a10b-8dd298d0e934.jpeg" />The initial prompt was:<br /></p><pre><code>1girl, standing, full body, turning around, looking back, long blonde hair, red jacket, blue jeans, park, grass, bench, concrete path, masterpiece, best quality, amazing detail, high definition</code></pre><p>Which could then be modified for the isolated subject:</p><p></p><pre><code>1girl, standing, full body, turning around, looking back, long blonde hair, red jacket, blue jeans, simple background, masterpiece, best quality, amazing detail, high definition</code></pre><p></p><p>and a background prompt:</p><p></p><pre><code>no humans, park, grass, bench, concrete path, masterpiece, best quality, amazing detail, high definition</code></pre><p>Masking out the subject becomes much easier with a simple background. Then it can be superimposed as a transparent subject over a fully detailed background. Even using a model that didn't understand "see-through body" we were still able to create synthetic training data so that it could learn it. Pretty neat!</p><p><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/cef0f475-fc54-4302-a06b-28dd0bfb2a67/width=525/cef0f475-fc54-4302-a06b-28dd0bfb2a67.jpeg" /></p><p>Before this I had no idea if this was a viable approach. The results, while not incredible, are better than I was expecting.</p>

See-through body | Experimental

<p>The v1 dataset trained on a network twice as large as before. Feedback is welcome.</p>

Files with SHA256: 4b6f258842bb68408bd0d6e080660b19d56e9281b890f45d742c66170434685a