<pre><code>Scanned VGA art is very expensive.</code></pre>When I made <a target="_blank" rel="ugc" href="https://civarchive.com/models/1239537?modelVersionId=1396980">the model for Sally</a>, I noticed she had an abnormally high inherent style for a LoRA of just 21 images (2 of which weren’t even from in-game). I liked the effect, so I decided to add some more training data to make a genuine style LoRA. Since Mortimer is pretty light on cutscenes I decided I’d try blending it with the relatively similar visually (but tonally opposite) Full Throttle from and The Dig from a similar release window for a total of 54 images. Curious to see how a style performs with a relatively limited dataset.Trigger: LucArtNi Optional: Flat color (helps force style in, but at cost of detail) Negative: Photo background (I don't know why, but it's needed to make the backgrounds less realisitic)Made with 54 pictures (19 reused from Sally, 10 more Mortimer pics with Sid, 17 Full Throttle, 8 The Dig). Curious how well a style will show up from this small a dataset. Rebel Assault 1, TIE Fighter and Dark Forces use a different style (higher detail, much less animation), and Outlaws is its own style, while several other LA titles from this era are starting to use prerendered 3D. Even the 2D cutscenes from the used games have some pretty obvious 3D I had to avoid putting in the dataset). Overall I'm less than satisfied with the results, it seems to do some things pretty well but have almost no effect on most things, but maybe you can pull something off with this anyways. Does anyone even want this for Pony? Might add some Herc's Adventures pics for a v2 if I ever do it. (Incidentally, if one wanted to pin a single name to the dataset, I think Anson Jew is the strongest influence here. He's the only one who worked on all three as far as I can tell.)

[IL, Experimental] Lucasarts 1995-96 Cutscene Style

IL_Lucasarts_9596_Cutscene_Style-000013.safetensors

<pre><code>Scanned VGA art is very expensive.</code></pre>When I made <a href="https://civitai.com/models/1239537?modelVersionId=1396980" rel="ugc noopener noreferrer nofollow" target="_blank">the model for Sally</a>, I noticed she had an abnormally high inherent style for a LoRA of just 21 images (2 of which weren’t even from in-game). I liked the effect, so I decided to add some more training data to make a genuine style LoRA. Since Mortimer is pretty light on cutscenes I decided I’d try blending it with the relatively similar visually (but totally opposite) Full Throttle from and The Dig from a similar release window for a total of 54 images. Curious to see how a style performs with a relatively limited dataset. Trigger: LucArtNi Optional: Flat color (helps force style in, but at cost of detail) Negative: Photo background (I don't know why, but it's needed to make the backgrounds less realisitic)Made with 54 pictures (19 reused from Sally, 10 more Mortimer pics with Sid, 17 Full Throttle, 8 The Dig). Curious how well a style will show up from this small a dataset. Rebel Assault 1, TIE Fighter and Dark Forces use a different style (higher detail, much less animation), and Outlaws is its own style, while several other LA titles from this era are starting to use prerendered 3D. Even the 2D cutscenes from the used games have some pretty obvious 3D I had to avoid putting in the dataset). Overall I'm less than satisfied with the results, it seems to do some things pretty well but have almost no effect on most things, but maybe you can pull something off with this anyways. Does anyone even want this for Pony? Might add some Herc's Adventures pics for a v2 if I ever do it. (Incidentally, if one wanted to pin a single name to the dataset, I think Anson Jew is the strongest influence here. He's the only one who worked on all three as far as I can tell.)

Files with SHA256: b542196704d382b5f85040fc8d70bbae682d159e4becb5d252039e21a28d72ad