linoyts/ideogram-4-turbotime

Ideogram 4 — TurboTime ⚡

Ideogram 4, Ideogram's open-weights text-to-image model, accelerated by Ostris's **TurboTime LoRA**: quality images in as few as 2–8 steps with no classifier-free guidance, versus the ~20–48 steps the base model needs.

Type a prompt, pick a step count (lower = faster), and generate.

Built on diffusers (Ideogram4Pipeline) on ZeroGPU. The nf4 weights are dequantized to bf16 at load and the TurboTime LoRA by Ostris is applied on top.