<div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/C5wZ2gU7j9E" start="0"></iframe></div>☕ One coffee = one happy creator: <a target="_blank" rel="ugc" href="https://buymeacoffee.com/the_frizzy1">https://buymeacoffee.com/the_frizzy1</a><hr />Native 2K Qwen Image 2.1 on a 12 GB card, with two paths in one graph. A QUALITY path at 40 steps with Comfy Kitchen attention, and a one-click TURBO path on Viggle's 6-step LoRA. Built from the official Text to Image template and retuned from my own benchmark: int8 model, int8 text encoder, 2048×2048, 40 steps instead of the template's 25.<h2 id="what-you-need">What you need</h2>✔ Custom nodes — None. Core ComfyUI only. ✔ ComfyUI version — 0.37 or newer — native Qwen Image 2.1, <code>ModelAttentionBackend</code>, <code>QwenImage21Cache</code>. 0.37.1 tested. ✔ VRAM — 12 GB — tested, not estimated ✔ System RAM — 3.0 GB peak on int8 (bf16 offloads and peaks at 10.6 GB) ✔ Tested on — RTX 3060 12 GB · 32 GB DDR4 · ComfyUI 0.37.1 (Comfy Desktop) ✔ Licence — Workflow MIT · Qwen Image 2.1 and the Viggle LoRA: Qwen Research License (non-commercial)<h2 id="models-to-download">Models to download</h2>Click a file name to download it. Put it in the folder shown, inside <code>ComfyUI/models/</code>.Required — 3 files, 17.38 GB<h3 id="diffusion_models">diffusion_models/</h3>File: qwen_image_2.1_int8_convrot.safetensors — 7.3 GB (Diffusion) Download: <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/diffusion_models/qwen_image_2.1_int8_convrot.safetensors">https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/diffusion_models/qwen_image_2.1_int8_convrot.safetensors</a>Put it here: ComfyUI/models/diffusion_models/<h3 id="text_encoders">text_encoders/</h3>File: qwen3vl_8b_int8_convrot.safetensors — 9.4 GB (Text encoder) Download: <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/text_encoders/qwen3vl_8b_int8_convrot.safetensors">https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/text_encoders/qwen3vl_8b_int8_convrot.safetensors</a>Put it here: ComfyUI/models/text_encoders/<h3 id="vae">vae/</h3>File: qwen_image_2.1_vae_bf16.safetensors — 0.68 GB (VAE) Download: <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/vae/qwen_image_2.1_vae_bf16.safetensors">https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/vae/qwen_image_2.1_vae_bf16.safetensors</a>Put it here: ComfyUI/models/vae/TURBO path only — 1 file.<h3 id="loras">loras/</h3>File: Qwen-Image-2.1-viggle-turbo-v0.2.1-6step-lora-r256.safetensors — 1.4 GB (Turbo LoRA) Download: <a target="_blank" rel="ugc" href="https://huggingface.co/Viggle/Qwen-Image-2.1-viggle-turbo/resolve/main/Qwen-Image-2.1-viggle-turbo-v0.2.1-6step-lora-r256.safetensors">https://huggingface.co/Viggle/Qwen-Image-2.1-viggle-turbo/resolve/main/Qwen-Image-2.1-viggle-turbo-v0.2.1-6step-lora-r256.safetensors</a>Put it here: ComfyUI/models/loras/Optional swaps, same repos, not in the workflow JSON: <code>qwen_image_2.1_bf16.safetensors</code> (14.2 GB), the bigger diffusion model, about 1.6× slower on 12 GB because it offloads to RAM · <code>qwen3vl_8b_w4a8.safetensors</code> (6.3 GB), smaller text encoder, same speed, different picture on the same seed.<h2 id="speed-on-my-rtx-3060">Speed on my RTX 3060</h2>This exact workflow, Comfy Desktop with SageAttention on, 2048×2048, warm run:<ul><li>QUALITY · 40 steps · Comfy Kitchen — 4 min 07 s time per image</li><li>TURBO · Viggle 6-step — 47 s time per image</li></ul>From the full benchmark, same machine, 2048×2048, 25 steps, same seed: PyTorch attention 3 min 57 s · SageAttention 3 min 00 s · Comfy Kitchen 2 min 40 s · Viggle 6-step 62 s on PyTorch and 50 s on Sage · Viggle 4-step 44 s.int8 against bf16, 2048×2048, 25 steps, PyTorch attention: 3 min 57 s against 6 min 10 s, model file 7.3 GB against 14.2 GB, peak system RAM 3.0 GB against 10.6 GB. bf16 does not fit in 12 GB, it offloads. Same-seed int8 and bf16 images measured SSIM 0.89–0.99 over 4 prompts.<h2 id="known-issues">Known issues</h2><ul><li>TURBO is softer in fine detail at 2K than the base model. Measured, and visible in 100% crops.</li><li>Seeds vary. Of two test seeds, TURBO double-exposed the subject's head on one and QUALITY printed the date line twice on one. Re-roll the seed.</li><li>Do not add Comfy Kitchen attention to TURBO. It ghosted the subject with the 6-step LoRA in my test. Leave TURBO on the ComfyUI default, which is Sage on Comfy Desktop.</li><li>TURBO sigmas depend on the size. Change the size, change the line — the table is in the workflow note.</li><li>The Viggle LoRA goes through ComfyUI's standard merging LoRA loader. Viggle's own unmerged-LoRA node was not used.</li><li>No peak VRAM figure was measured for this one, only peak system RAM. So the 12 GB number is "it runs on a 12 GB card", not a nvidia-smi reading.</li></ul><h2 id="get-the-workflow">Get the workflow</h2><ul><li>GitHub: <a target="_blank" rel="ugc" href="https://github.com/Thefrizzy1/The_frizzy1-Workflows/tree/main/workflows/qwen/image-2.1-12gb">https://github.com/Thefrizzy1/The_frizzy1-Workflows/tree/main/workflows/qwen/image-2.1-12gb</a></li><li>Hugging Face mirror: <a target="_blank" rel="ugc" href="https://huggingface.co/The-frizzy1/The_frizzy1-Workflows/tree/main/workflows/qwen/image-2.1-12gb">https://huggingface.co/The-frizzy1/The_frizzy1-Workflows/tree/main/workflows/qwen/image-2.1-12gb</a></li></ul>Second video, the settings one: <a target="_blank" rel="ugc" href="https://www.youtube.com/watch?v=-yJvQ3lxWsE&amp;list=PLG_da2oqTx7k">https://www.youtube.com/watch?v=-yJvQ3lxWsE&amp;list=PLG_da2oqTx7k</a>More videos on this stuff: <a target="_blank" rel="ugc" href="https://www.youtube.com/playlist?list=PLG_da2oqTx7k">https://www.youtube.com/playlist?list=PLG_da2oqTx7k</a><hr />

Qwen Image 2.1 12GB — int8 Native 2K Text to Image, 40-step QUALITY + Viggle 6-step TURBO LoRA

v2.0.0. The README points at <code>changelog.md</code> without saying what changed, so there is no cue to lift here.

Files with SHA256: ab1e7da98abcac3f6cdccf0dea72f79e3812e8d398d4e56ab231985485e7cbcf