Sulphur 2 Base is an uncensored, community-built video generation model on top of LTX 2.3. It runs text-to-video and image-to-video natively and supports every other LTX 2.3 format, with an optional prompt enhancer and distill LoRA in the same release.
Originally released by <a target="_blank" rel="ugc" href="https://huggingface.co/SulphurAI/Sulphur-2-base">SulphurAI</a> on Hugging Face. All credit for the model goes to the SulphurAI team and contributors below. Civitai is hosting a mirror so creators can run it on-site - please head to the <a target="_blank" rel="ugc" href="https://huggingface.co/SulphurAI/Sulphur-2-base">original repo</a> for weights, updates, and to support the project directly.
<h3 id="built-by">Built by</h3>
<ul>
 <li>SulphurAI - <a target="_blank" rel="ugc" href="https://huggingface.co/SulphurAI/Sulphur-2-base">Hugging Face</a> | <a target="_blank" rel="ugc" href="https://discord.gg/GSXJhKZ9V">Discord</a> | <a target="_blank" rel="ugc" href="https://ko-fi.com/fusioncow">Ko-fi</a></li>
 <li><a target="_blank" rel="ugc" href="https://huggingface.co/TenStrip">TenStrip</a> - testing and merging (<a target="_blank" rel="ugc" href="https://huggingface.co/TenStrip/LTX2.3-10Eros">LTX2.3-10Eros i2v merge</a>)</li>
 <li><a target="_blank" rel="ugc" href="https://huggingface.co/silveroxides">silveroxides</a> (<code>@s1lv3rc01n</code>) - testing, merging, quantizing</li>
 <li><code>@mov7162</code> - Musubi Tuner guidance</li>
 <li>Two anonymous funders who made the original Sulphur and Sulphur 2 possible</li>
 <li>Many other community contributors</li>
</ul>
<h3 id="native-t2v-and-i2v-on-ltx-23">Native t2v and i2v on LTX 2.3</h3>
Sulphur 2 inherits the LTX 2.3 pipeline end to end, so anything an LTX 2.3 workflow can do (t2v, i2v, the rest of the 2.3 formats) works here without extra glue. No converter nodes, no sidecar checkpoints.
<h3 id="versions-on-civitai">Versions on Civitai</h3>
<ul>
 <li>Dev - <code>fp8mixed</code> build for local runs that need VRAM headroom</li>
 <li>Distilled - the included distill LoRA, ideal as the recommended starting point</li>
</ul>
Pair the Dev version with the Distilled LoRA for the lightest setup. Existing workflows still reference <code>sulphur_final</code>. Use the LoRA or the full model, never both at the same time. The HF release also ships a <code>bf16</code> build, which is not currently mirrored on Civitai.
<h3 id="built-in-prompt-enhancer">Built-in prompt enhancer</h3>
A multimodal prompt enhancer ships with the model and is designed to drop straight into LM Studio:
<ol>
 <li>Open your LM Studio model folder.</li>
 <li>Create <code>Sulphur/promptenhancer/</code>.</li>
 <li>Drop the <code>.gguf</code> and <code>mmproj</code> files inside.</li>
</ol>
LM Studio will pick it up on next launch. No system prompt is required, send the text (and an optional image) and it returns an enhanced prompt.
<h3 id="uncensored-by-design">Uncensored by design</h3>
Sulphur 2 is intentionally unfiltered for adult and creative work. Generations on Civitai still respect the site's content rules.
<h3 id="links">Links</h3>
<ul>
 <li>Source: <a target="_blank" rel="ugc" href="https://huggingface.co/SulphurAI/Sulphur-2-base">https://huggingface.co/SulphurAI/Sulphur-2-base</a></li>
 <li>Discord: <a target="_blank" rel="ugc" href="https://discord.gg/GSXJhKZ9V">https://discord.gg/GSXJhKZ9V</a></li>
 <li>Support the project: <a target="_blank" rel="ugc" href="https://ko-fi.com/fusioncow">https://ko-fi.com/fusioncow</a></li>
</ul>

Files with SHA256: 41c999575859c528ff108022246a5524960a778c18742696971c9b0aadb4f70f