This is a melting and glowing style. The prompt should start with <code>melty</code> and can include anything else you can imagine. It will not turn everything into a puddle of slime. Tends to make a lot of spheres and pools of water.This was trained on a micro dataset of 10 images generated by Flux with a 16:9 aspect ratio suitable for desktop backgrounds. This took about 60 minutes to train and seems to work with a strength from 1.0 - 1.5, so it is probably a little bit under-trained. This is another test for a different training method, this time using a very small dataset and captions that each have 2 words, <code>melty [something]</code>.<h2 id="concepts-wop9lv99n">Concepts</h2><ul><li>melty (melting)<ul><li>planets</li><li>metal</li><li>fire</li><li>stars</li><li>energy</li></ul></li><li>glowing (not captioned)</li></ul><h2 id="example-prompts-tajlhgift">Example Prompts</h2><ul><li>melty planets colliding in a stellar explosion</li><li>melty cybernetic diamonds suspended in the mainframe</li></ul><h2 id="training-parameters-ppddcpwvo">Training Parameters</h2><pre><code>params:
 d_coef: 2
 learning_rate: 1.0
 learning_scheduler: cosine_with_restarts
 max_epochs: 15
 network_alpha: 32
 network_dim: 32
 noise_offset: 0.1
 num_cycles: 5
 optimizer_type: prodigy
 optimizer_args: |
 "decouple=True" "weight_decay=0.1" "betas=0.9,0.999" "use_bias_correction=False" "safeguard_warmup=False"
 save_every_n_steps: 40
 sample_every_n_steps: 40
 timestep_sampling: shift
 weight_decay: 0.1

samples: flux_melty_test</code></pre>and<pre><code>{
 "datasets": [{
 "resolution": 1280,
 "enable_bucket": true,
 "min_bucket_reso": 768,
 "max_bucket_reso": 1280,
 "bucket_reso_steps": 256,
 "bucket_no_upscale": true,
 "batch_size": 6,
 "subsets": [{
 "image_dir": "/mnt/training/flux/flux-melty-test",
 "caption_extension": ".txt",
 "num_repeats": 5
 }]
 }]
}</code></pre>This is the same configuration shown in <a target="_blank" rel="ugc" href="https://civarchive.com/models/1079099">my Flux Nude/NSFW Training Base model</a>, but this LoRA does not include nudity. I wanted to test that configuration without a base LoRA and training from Flux itself. This is the result. Please let me know what you think of it.
<hr />

Created by <a rel="ugc" href="https://thalis.dev">Thalis AI</a> — dark fantasy, cosmic horror, and the spaces between.

Support on <a rel="ugc" href="https://ko-fi.com/thalisai?utm_source=civitai&amp;utm_medium=model_desc">Ko-fi</a> — every tip keeps the radium gems glowing.
Join on <a rel="ugc" href="https://www.patreon.com/ThalisAI?utm_source=civitai&amp;utm_medium=model_desc">Patreon</a> — early access, prompt packs, and the deeper corridors.
<a rel="ugc" href="https://thalis.dev/models/?utm_source=civitai&amp;utm_medium=model_desc">Full catalog</a> — 100+ LoRAs for Flux, Illustrious, and more.

Melty Style (Melting and Glowing)

flux-melty-test-v02.safetensors

flux-melty-test-dataset.zip

v2.0

This is a melting and glowing style. The prompt should start with <code>melty</code> and can include anything else you can imagine. It will not turn everything into a puddle of slime. Tends to make a lot of spheres and pools of water.This was trained on a micro dataset of 10 images generated by Flux with a 16:9 aspect ratio suitable for desktop backgrounds. This took about 60 minutes to train and seems to work with a strength from 1.0 - 1.5, so it is probably a little bit under-trained. This is another test for a different training method, this time using a very small dataset and captions that each have 2 words, <code>melty [something]</code>.<h2 id="concepts-wop9lv99n">Concepts</h2><ul><li>melty (melting)<ul><li>planets</li><li>metal</li><li>fire</li><li>stars</li><li>energy</li></ul></li><li>glowing (not captioned)</li></ul><h2 id="example-prompts-tajlhgift">Example Prompts</h2><ul><li>melty planets colliding in a stellar explosion</li><li>melty cybernetic diamonds suspended in the mainframe</li></ul><h2 id="training-parameters-ppddcpwvo">Training Parameters</h2><pre><code>params:
 d_coef: 2
 learning_rate: 1.0
 learning_scheduler: cosine_with_restarts
 max_epochs: 15
 network_alpha: 32
 network_dim: 32
 noise_offset: 0.1
 num_cycles: 5
 optimizer_type: prodigy
 optimizer_args: |
 "decouple=True" "weight_decay=0.1" "betas=0.9,0.999" "use_bias_correction=False" "safeguard_warmup=False"
 save_every_n_steps: 40
 sample_every_n_steps: 40
 timestep_sampling: shift
 weight_decay: 0.1

samples: flux_melty_test</code></pre>and<pre><code>{
 "datasets": [{
 "resolution": 1280,
 "enable_bucket": true,
 "min_bucket_reso": 768,
 "max_bucket_reso": 1280,
 "bucket_reso_steps": 256,
 "bucket_no_upscale": true,
 "batch_size": 6,
 "subsets": [{
 "image_dir": "/mnt/training/flux/flux-melty-test",
 "caption_extension": ".txt",
 "num_repeats": 5
 }]
 }]
}</code></pre>This is the same configuration shown in <a target="_blank" rel="ugc" href="https://civitai.com/models/1079099">my Flux Nude/NSFW Training Base model</a>, but this LoRA does not include nudity. I wanted to test that configuration without a base LoRA and training from Flux itself. This is the result. Please let me know what you think of it.

Files with SHA256: 25ef45437b7dfec1abdeb2b270edf40d976c36cb81d303748bbc4449d8bd9115