import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("Bl4ckSpaces/Project-Cosmos-V1.5", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]Project-Cosmos-V1.5 (Dual-Mode TIES)
Cosmos V1.5 uses an advanced Dual-Mode Delta Gating technique to surgically separate Anatomy from Style.
The "Advanced" Logic
Instead of simple block swapping, we analyze the Magnitude of Change (Delta) for every single parameter.
Anatomy Layers (Input/Middle) -> "Anti-TIES" Mode:
- We inspect every update from NoobAI.
- If the update is Large (Structural Change/Bone Movement) -> REJECTED.
- If the update is Small (Micro-Texture/Skin Pore) -> ACCEPTED.
- Result: 100% IllumiYume Bone Structure + NoobAI Skin Texture.
Style Layers (Output) -> "Pro-Style" Mode:
- We force a 95% transition to NoobAI's rendering engine using high-precision blending.
- Result: Clean, soft lighting and high CFG stability.
Recommended Settings
- Sampler: Euler a / DPM++ 2M SDE
- CFG: 5.0 - 9.0
- Steps: 28+
- Downloads last month
- 9