[SD 2.1 768][Checkpoint]fking_scifi_v2 v2.1f
<h1>fking_scifi_v2</h1><p><strong>A fking StableDiffusion model</strong></p><p>Trained on 666 images, hand captioned, general, but limited sci-fi model. It has been trained on armour concepts, futuristic backgrounds, androids and robots, as well as some fantasy stuff, like werecreatures.</p><p></p><p><strong>Additional Notes</strong></p><p>Be sure to grab to <code>YAML</code> file too, and place it in the same directory as the downloaded model. It must be named exactly the same, except for the extension, i.e, <code>fking_scifi_v2.ckpt</code> and <code>fking_scifi_v2.yaml</code>.</p><p></p><h2><strong>Usage</strong></h2><p>Primary trigger word is <code>fking_scifi_v2</code>, but you can also use <a target="_blank" rel="ugc" href="https://fking.gay/u/FZ8tcjp">this to help</a> guide your prompting.</p><p></p><h2><strong>Minor Update v2.1f</strong></h2><p>I've decided to release a small update to this model that is a merge between <code>fking_scifi_v2</code> and some of the various test checkpoints of the upcoming <code>fking_scifi_v3</code>, you still only need to define the trigger word <code>fking_scifi_v2</code> as those test checkpoints were trained with that trigger. This <code>v2.1</code> model is a bit harder to control, but in my opinion, makes better photo-realistic creatures. Because it is a merge, its very possible that it "forgot" previous concepts, and for that, I apologize. Feel free to use either version, but if you are wanting advise, or help, please state which model version you are using so that I can assist you better.</p><p><em>Just a note, you may see some of the eyes wonky in the previews, but I wanted to show off only what the model is capable of, no face fixes or other embeds. Best of luck guys!</em></p><p><em>PPS. Mermaids are seriously hard to generate with base SD and this model, don't be discouraged, that sample was cherry picked (but not edited, or post-processed).</em></p><p></p><h2><strong>Future Updates</strong></h2><p>Next version will include a wider encyclopedia of concepts, including, wizards, elfs, druids, more dragons, skeleton creatures, aliens, demons, golems, different backgrounds and environments, and many more. Trying to include POC and female characters as well, by default male characters will have no gender modifier, but female characters can be invoked with <code>female</code>, <code>woman</code>, and/or <code>lady</code>. This is because, in my opinion, SD is already heavily biased towards men, and I do not want to overtrain the word.</p><p></p><p><strong>Progress</strong></p><ul><li><p><strong>[<em>DONE*</em>]</strong> Gather concept images</p></li><li><p><strong>[<em>INCOMPLETE</em>]</strong> Create crops of concept images</p></li><li><p><strong></strong> Caption concept images</p></li><li><p><strong></strong> Test learning rates/ability to build upon previous checkpoint without losing, or overtraining previous concepts</p></li><li><p><strong></strong> Training</p></li><li><p><strong></strong> Testing/Fine-tuning</p></li><li><p><strong></strong> Finalize</p></li><li><p><strong></strong> Publish</p></li></ul><p><em>*I keep finding more, and just can't help myself.</em></p><p></p><p>Just a note, but I probably worked on and off for several weeks on the V2 version of this model, and several small attempts at V1, but I plan to take my time with V3, so I'm sorry if anyone is impatient.</p><p></p><p><strong>V3 Samples</strong></p><p>I have been testing different learnings on a smaller dataset of my v3 concepts and I have had great results so far, <a target="_blank" rel="ugc" href="https://fking.gay/u/QadR1xt">here are some previews of v3</a>.</p><p></p><h2>Setbacks</h2><p>Only a small one, my x768 version looked like ass, and I decided to scrap it completely, only wasted 18 hours. New model will be continued on x512 images instead as I seem to have a better grasp at training on that size.</p><p></p><h2>StableDiffusion 1.x</h2><p>After testing and much debate, I have decided not to release a 1.x model. Yea I know, I'm sorry, but I am not satisfied with the results, and I do not want to share something that does not meet my standards and expectations. Long-live StableDiffusion 2.1</p><p></p><h2><strong>Contributions</strong></h2><p>If you would like to contribute, either concept images, or help with the laborious process of captioning images, message me on Discord, <code>raekor.#0768</code>.</p><p></p><p><strong>Remixes/Merges</strong></p><p><em>I'd prefer not to have merges of this model redistributed, but I am not going to report you, and I am not going to bitch about it, just make them of a higher quality.</em></p><p></p><p><strong><em>Please remember to like or rate model. Happy prompting!</em></strong></p>
页:
[1]