<h1 id="overview">Overview</h1>An experimental LoRA for some female height growth. Even state-of-the-art models struggle with this concept, so don't expect mind blowing results. Playing around with it was a lot of fun though.Use the trigger and try to describe her final form with <code>after growing into a giantess...</code>.For more have a look at the example images/videos.<h3 id="h3-v1">H3 - V1</h3>Use the trigger <code>gtsgrw</code> or <code>Her height increases</code> then describe what you want. Describe the body type if you want the woman to stay slim or become bigger too. Describe relations to her surroundings are other people. Use lower strengths at first, start with 0.3.There are also separate LoRAs for only the motion/growth part and the static giantess part (<code>separate_motion</code>/<code>separate_static</code>). You can play around with these for more control. The <code>separate_static</code> can produce better giantess content if used with T2V.<ul><li>Training model: <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/MiniMax-H3">H3 FL2VA</a></li><li>LoRA strength: 0.3-0.9</li><li>Sampler: euler / res_multistep / res_2s+beta57</li><li>Steps: 20+</li></ul><pre><code>gtsgrw, A realistic video of a young woman sitting on a couch.

She looks shocked as her height increases. Her giant legs stretch across the room Her head bumps into the ceiling, cracking it. Plaster and dust falling down.

After becoming a giantess, her clothes are tight and her body is curvy.

She says: "Ouch!", after hitting the ceiling.</code></pre><h3 id="ltxv-2.3-sulphur-v1">LTXV 2.3 Sulphur - V1</h3>Does work on LTX 2.5 with limitations.Intended to be used with a I2V LTX 2.3 10Eros workflow. T2V will yield unpredictable results.Not enough clips in the dataset to train LTX properly, yet. Still releasing this because it's fun to experiment with 😏<ul><li>Training model: <a target="_blank" rel="ugc" href="https://huggingface.co/SulphurAI/Sulphur-2-base">LTX 2.3 Sulphur</a></li><li>LoRA strength: 0.5-1.3</li><li>LoRA strength high-res: 0.15-0.5</li><li>Sampler: euler</li><li>CFG: 3.0</li><li>Steps: 20+</li></ul><pre><code>grwth, A realistic 7-second live-action video of a woman standing in a narrow alley, wearing a t-shirt and black leggings. her height increases rapidly and dramatically. after growing to a giantess, her head towers above the roofs.

With a distressed tone she says: "Oh no... NOT NOW!", while looking down at her giant body. Her hands reach down to rest on the edges of the roofs.

Quiet ambiance.</code></pre><h3 id="wan-2.2-i2v-v1">WAN 2.2 I2V - V1</h3>High-noise model is required, while the low-noise may be omitted or swapped out with other detailers or low-noise models. I also recommend using <code>rife49</code> frame interpolation to double the frames, then set the fps to 24 so get a longer, smoother video.<ul><li>Training model: <a target="_blank" rel="ugc" href="https://huggingface.co/Wan-AI/Wan2.2-I2V-A14B">WAN 2.2 I2V</a></li><li>LoRA strength: 0.5-1.0</li><li>Sampler: euler / euler a</li><li>Steps: 8+</li><li>Using the <a target="_blank" rel="ugc" href="https://github.com/VraethrDalkr/ComfyUI-TripleKSampler">TripleKSampler</a> or higher steps can greatly improve the animation quality, if you're using speedup LoRAs. For reference, all showcase videos have been generated with 8 steps, but using the full model or the triple sampler will improve the quality drastically.</li></ul><pre><code>grwth, A woman sitting on a couch. Her height increases. After growing into a giantess, she towers over the living room, her long legs and giant feet reach forward.</code></pre>

💁‍♀️ Height Growth

h3_height_growth_v1.safetensors

h3_growth_separate_motion_v1.safetensors

Files with SHA256: 1d2eba68f9bbac9288e846929f0f8b195c72dd863362a5db4795c7a98ebf2258