MOSSKVA — MOSS-Image SFT Merge (INT8 ConvRot). ADVANCED HANDCRAFTED MERGE based on MOSS-Image SFT (5B DiT — Diffusion Transformer with Flow Matching), blending three specialized LoRAs for anatomy and raw amateur realism. MERGE RECIPE (calibrated to avoid overbaking): MysticXXX Moss V1 (NSFW & Anatomy) 0.60 | Lenovo Moss (Primary Amateur Photorealism) 0.50 | NiceGirls V2 Moss (Secondary Photorealism) 0.30. FEATURES: Hyper-realistic skin texture eliminating the plastic 3D render look — Natural amateur lighting: smartphone flash, low light, authentic colors — Uncensored anatomical understanding without forcing explicit content — Keeps casual dressed portrait capability. QUANTIZATION (INT8 ConvRot — 5.04 GB): TensorWiseINT8Layout with regular Hadamard ConvRot rotation (group 256) — Fidelity: practically identical to the original BF16 (10.05 GB), CosSim = 0.9933 — Direct GPU quantization using native ComfyUI/comfy-kitchen layouts. REQUIRED FILES IN COMFYUI: Text Encoder (models/text_encoders/): qwen3.5_9b_w4a8_gptq.safetensors (CLIPLoader, type: moss_image) — VAE (models/vae/): flux2-vae.safetensors (VAELoader to VAEDecode) — Empty Latent: EmptyFlux2LatentImage (channels=128, downscale=16x). RECOMMENDED SETTINGS (Standard SFT): Steps 30-50 (ideal 35) — CFG Scale 3.0-4.0 (default 3.0) — Sampler: euler or dpmpp_2m_sde — Scheduler: sgm_uniform or beta — Resolution 1024x1024 / 896x1152 — Model Sampling Shift 5.0 (AuraFlow / Flow) — Average time on RTX 5060 Ti: 8-15 sec per image. PROMPT TRIGGERS FOR AMATEUR REALISM: "candid amateur phone photo, casual selfie taken with an iPhone/Samsung" — "dim room lighting, direct flash photography, authentic harsh shadows" — "natural uneven skin texture, subtle skin imperfections, visible pores, no makeup" — "spontaneous pose, slightly grainy film look, messy background, candid snapshot". AVOID in positive prompt: masterpiece, 8k, hyperrealistic, octane render, flawless skin, cinematic lighting (forces artificial smooth video-game skin). VERSIONS: mosskva 1.0 (INT8 ConvRot, this version, 5.04 GB) — mosskva 1.0 (W6A8 ConvRot, 4.21 GB) — mosskva 1.0 (NVFP4 Hybrid, 4.20 GB). Additional Turbo DMD2 versions (8 steps, CFG 1.0) also available in this series. LICENSE: allows use without credit, allows sharing merges, allows different licenses on merges. Commercial use allowed. Not intended to depict actual persons.
Description
WHAT THIS VERSION IS:
mosskva nvfp4 is the NVFP4 Hybrid quantized variant (4.20 GB) of the mosskva 1.0 Standard SFT merge, built on MOSS-Image SFT (5B DiT with Flow Matching). It uses a HYBRID quantization layout: TensorCoreNVFP4Layout (E2M1 with MSE-optimal block scales) on early blocks (1-20) with low dynamic range, plus TensorWiseINT8Layout (ConvRot 256) on input block (0), intermediate output projections, and late blocks (21-41) where MOSS-Image activations peak. This hybrid design eliminates the patchy artifacts of pure NVFP4 while preserving fine texture integrity. Cosine similarity to INT8 ConvRot reference: 0.974+. File: moss_image_sft_mystic_lenovo_nicegirls_nvfp4.safetensors (4.20 GB).
⚡ OPTIMIZED FOR RTX 50-SERIES (BLACKWELL):
This version leverages native Tensor Core FP4 (E2M1) execution on NVIDIA Blackwell architecture (RTX 50-series). The NVFP4Layout uses hardware-accelerated FP4 tensor cores with MSE-optimized block scales — delivering maximum throughput on RTX 5060/5070/5080/5090. On pre-Blackwell GPUs (RTX 30/40 series), it falls back to emulated FP4 via INT8 kernels with slight overhead. For RTX 50xx users, this is the RECOMMENDED Standard SFT variant — smallest size (4.20 GB), fastest inference, native hardware support.
HOW TO LOAD IN COMFYUI:
- Loader: UNETLoader (standard) — loads the .safetensors directly as a checkpoint
- Text Encoder: qwen3.5_9b_w4a8_gptq.safetensors (models/text_encoders/) via CLIPLoader with type="moss_image"
- VAE: flux2-vae.safetensors (models/vae/) via VAELoader
- Latent: EmptyFlux2LatentImage (channels=128, downscale=16x)
MERGE RECIPE (fused in Float32 over BF16 base, then GPU-quantized to NVFP4 Hybrid):
- MysticXXX Moss V1 (NSFW & Anatomy): strength 0.60
- Lenovo Moss (Amateur Photorealism): strength 0.50
- NiceGirls V2 Moss (Secondary Photorealism): strength 0.30
STANDARD SFT SETTINGS (30-50 steps, CFG 3.0-4.0):
- Steps: 30-50 (ideal 35)
- CFG: 3.0-4.0 (default 3.0)
- Sampler: euler or dpmpp_2m_sde
- Scheduler: sgm_uniform or beta
- Resolution: 1024x1024 / 896x1152
- ModelSampling Shift: 5.0 (AuraFlow / Flow)
- Avg time: ~8-15s/image (RTX 5060 Ti), FASTER on RTX 50-series via native FP4 tensor cores
QUANTIZATION TECHNICAL DETAILS:
- Layout: Hybrid TensorCoreNVFP4Layout (blocks 1-20) + TensorWiseINT8Layout ConvRot 256 (blocks 0, 21-41, intermediate projections)
- Quantization: Per-block MSE-optimal scale search on NVFP4 blocks; per-group MSE (group 16) on INT8 ConvRot blocks
- File: moss_image_sft_mystic_lenovo_nicegirls_nvfp4.safetensors (4.20 GB)
- Requires: qwen3.5_9b_w4a8_gptq.safetensors (text encoder), flux2-vae.safetensors (VAE)
TRIGGER WORDS (recommended):
candid amateur phone photo, smartphone photo, direct flash, authentic harsh shadows, natural uneven skin texture, subtle skin imperfections, visible pores, no makeup, spontaneous pose, slightly grainy, messy background, candid snapshot
IMPORTANT: Do NOT use words like "masterpiece, 8k, hyperrealistic, octane render, flawless skin, cinematic lighting" — these force artificial smooth skin and defeat the amateur realism aesthetic.