realistic-vision-cfg-onnx

Realistic Vision V6 (CFG, full quality) — ONNX

ONNX export of SG161222/Realistic_Vision_V6.0_B1_noVAE paired with stabilityai/sd-vae-ft-mse. No distillation LoRA — this is the full, non-distilled UNet driven with classifier-free guidance. SD 1.5 architecture, 512×512 native, Euler scheduler, CFG ≈ 7.5, ~25 steps.

This is the quality counterpart to the 4-step Hyper export. The Hyper variant is distilled for fast, CFG-free, 1–4 step sampling; it's great for previews and batch work but caps fidelity and prompt adherence. This non-distilled export, run with classifier-free guidance, a negative prompt, and a normal step budget, recovers the sharp, prompt-faithful, clean-tone output people expect from Realistic Vision — at a higher per-image cost (the UNet runs twice per step, over ~25 steps).

Realistic Vision V6 is the photorealistic-portrait flagship of the SD 1.5 ecosystem. Trained on a narrow distribution (people, portraits, photography aesthetics), which is exactly why it's more stable across seeds than base SD 1.5 for those subjects.

Heads-up: Realistic Vision is more NSFW-permissive than the other variants in this collection. Pair with content filters if that matters for your application.

Converted artifact. Training credit: SG161222 (Realistic Vision), Stability AI (sd-vae-ft-mse).

What this repo contains

model_index.json
feature_extractor/
scheduler/
text_encoder/
tokenizer/
unet/                   # RV6 UNet, non-distilled (no LoRA)
vae_decoder/            # sd-vae-ft-mse (RV6 ships without VAE — paired here)
vae_encoder/

How it was produced

1. Load SG161222/Realistic_Vision_V6.0_B1_noVAE via diffusers.
2. Replace the (missing) VAE with stabilityai/sd-vae-ft-mse — the SD 1.5 community-standard fine-tuned VAE.
3. optimum-cli export onnx (no LoRA fusion step).

Exported at FP32 — the SD 1.5 VAE is fp16-fragile (it overflows and posterizes), so the quality export stays full precision.

Toolchain: optimum 1.24.0, diffusers 0.31.0, transformers 4.45.2, torch 2.4.x (CUDA 12.4). Conversion script: scripts/export-realistic-vision-cfg.ps1.

Inference notes

Classifier-free guidance runs the UNet twice per step (conditional + unconditional) and combines them as uncond + guidance · (cond − uncond). The negative prompt only takes effect when guidance > 1.

License

CreativeML OpenRAIL-M (SD 1.5 + Realistic Vision). License files included. By using this model you accept those terms.