Sulphur — Uncensored LTX-2.3 Video Generation
Parameters
Generation Mode
Text to Video
Image to Video
Prompt (cinematographer-style works best)
Negative Prompt
Resolution (stage 1 — output is 2x)
512x512 - 1:1 (→1024×1024)
768x512 - 3:2 (→1536×1024)
832x480 - 16:9 (→1664×960)
Frame Count
25 frames - ~1 second
49 frames - ~2 seconds
121 frames - ~5 seconds
Inference Steps: 30
30 default. With distill LoRA, 8–15 also works.
Guidance Scale: 3.0
3.0 default
Advanced Settings
Seed (optional)
Note:
Sulphur is a 22B LTX-2.3 fine-tune (BF16 → FP8 at runtime). Cold-start load is ~60–90 s; warm requests run in ~5–60 s depending on resolution and frame count. ~40–50 GB VRAM peak.
Generate Text-to-Video
Generated Video
Enter a prompt to generate a video