Sulphur 2, built on the LTX ecosystem, is a text-to-video and image-to-video model focused on realistic motion and cinematic-quality AI video generation.
Sulphur 2 is a community-released video generation model with 9B parameters. It gives you creative control. Fine-tuned on about 125,000 video clips, it builds on the LTX 2.3 base. The package includes open weights, distill LoRAs, ComfyUI workflows, and a local prompt enhancer. There’s no hosted SaaS layer between you and your output. Everything runs on your hardware, at your pace.
Turn a natural-language prompt into continuous, stable motion. Sulphur 2 base builds on LTX 2.3's temporal consistency design. It maintains coherence in subjects, lighting, and camera movement throughout the clip. There’s no flickering or drifting.
Animate any still image into a fluid, coherent shot. Sulphur 2 makes your static visuals come alive. You can use concept art, character renders, or storyboard frames. No need for manual frame-by-frame rekeying!
You can find the complete 9B-parameter base model on Hugging Face at SulphurAI/Sulphur-2-base. It includes distill LoRAs that greatly reduce sampling steps. This makes quick testing on local consumer hardware easy and practical.
Official ComfyUI workflows ship with the model. Drop the JSON into ComfyUI, load the weights, and your text-to-video and image-to-video pipeline is live. The bundled local prompt enhancer boosts your outputs. You don’t need prompt engineering skills.
Sulphur 2 (model ID: SulphurAI/Sulphur-2-base) is a community-distributed video generation model with 9 billion parameters. It runs on LTX 2.3 and was released on May 3, 2026. This model supports both text-to-video and image-to-video generation. It comes as a complete bundle, including open weights, distill LoRAs, ComfyUI workflows, and a prompt enhancer.
LTX 2.3 is the upstream base model. Sulphur 2 is a community fine-tune. It builds on the original and is trained on about 125,000 video clips. Plus, it's packaged as a ready-to-run local setup. If you already run LTX 2.3, integrating Sulphur 2 requires minimal changes to your existing workflow.
A discrete GPU with 24–32 GB of VRAM handles the base safetensors cleanly. 32 GB or more is recommended for higher resolutions or longer clip durations. Apple Silicon and lower-VRAM GPUs can run distilled or quantized variants through ComfyUI, though per-clip generation times will be longer.
Commercial use is governed by the Sulphur 2 model card license and the upstream LTX 2.3 terms at the time of download. Review both documents carefully before publishing outputs in any commercial context. Nothing here constitutes legal advice.
