The plain Qwen Image 2.1 text-to-image graph: your prompt, your canvas, an optional LoRA stack, a PNG. Core Qwen nodes plus my free AusBoss nodes, nothing else.
New video: I threw a ton of edits at Qwen Image 2.1, FLUX 3 Klein and FIBO 2
The first run renders the night bus-stop portrait from the gallery at 1536 × 1536 in 25 steps, in about 13 seconds on an RTX 5090.
What's inside
LoRA stack: starts empty. Add only adapters made for Qwen Image 2.1; trigger words of the rows you switch on join your text by themselves.
Canvas: Latent Size picks the size (1536 × 1536 here) and drives the core empty latent. Keep its 32-pixel snap.
Prompt: a Text node into Text Encode Qwen Image 2.1, which handles the positive and negative conditioning together.
Seed: starts on Random, so every run after the first gives a new picture. Switch it to Fixed to keep one.
Save: a PNG with the workflow embedded, plus a render timer and the Workflow Note card with download buttons.
Quick start
Update ComfyUI. You need a build with Text Encode Qwen Image 2.1 (mid-September 2026 or newer; I tested on 0.37.0).
Install AusBoss nodes from ComfyUI-Manager (search "AusBoss", registry id
ausboss-nodes) and restart. You need 2.2.0 or newer. Manager installs the newest.Drag the workflow in and download the three files from the card into the folders it names. No input image needed.
Write a prompt and queue.
Models
qwen_image_2.1_int8_convrot.safetensors →
models/diffusion_models· 7.26 GBqwen3vl_8b_int8_convrot.safetensors →
models/text_encoders· 9.35 GBqwen_image_2.1_vae_bf16.safetensors →
models/vae· 676 MB
BF16 versions are on the same page: Comfy-Org/Qwen-Image-2.1
Custom nodes
Only ComfyUI-AusBoss, 2.2.0 or newer (Manager installs it): free and open source. Text Encode Qwen Image 2.1 and the sampler are core ComfyUI. You don't need a Qwen wrapper pack.
Settings that worked, and what didn't
Sampling: 25 steps, CFG 1, Euler / simple, denoise 1: the starting values of the official template. 40 steps gives a more complete render. The negative prompt has no effect at CFG 1.
Canvas: 1024 × 1024 is a fine start; try 2048 × 2048 for a larger render. The sample runs at 1536 × 1536; the rooftop shot in the gallery uses 1152 × 2048.
Transparent output: ask explicitly for an RGBA image, a transparent background and an alpha channel, and keep PNG selected so the transparency survives saving.
Speed
Measured on an RTX 5090 (32 GB) with the INT8 files and the models loaded: about 13 s at 1536 × 1536 or 1152 × 2048, 25 steps.
I haven't tested smaller cards yet. ComfyUI offloads what doesn't fit, so expect it to run slower rather than fail.
About the gallery
All three portraits are straight outputs of this workflow; drag a PNG into ComfyUI to load its exact settings. The night bus stop is the default prompt (seed 7812); the rooftop at golden hour (seed 6311) and the photo booth (seed 6101) are other prompts. The cover is the bus-stop render plus a screenshot of the workflow's own Prompt card.
License & credits
Qwen Image 2.1 is © Alibaba Qwen, released under the Qwen Research License: non-commercial use only; commercial use needs a separate license from Qwen. The workflow and the AusBoss nodes are free.
Changelog
v1.0 (2026-09-25): first release.
Questions or bugs: GitHub issues or the comments here. Post what you make with it; I read everything. Videos on YouTube @ausbossdev. I post new workflows on X @Zanzibased and GitHub.
Description
First release.
FAQ
Comments (2)
This workflow was created solely to promote some moronic node pack from the manager; everything it contains has long had standard, well-established alternatives (seed, image size, prompt input). I really dislike these gimmicky workflows that add nothing new and just clutter up ComfyUI. Once the nodes are swapped for standard ones, the workflow works fine, but it’s completely average - nothing special at all.
fair point. you're right that core covers seed, size and prompts, and the workflow runs fine with those. i started out making nodes with features i wanted and didn't see anywhere, like scrubbing lora strengths and seeing them as bars, or the crop rotate pad nodes. i probably got carried away with a seed node that doesn't add much. appreciate the feedback, i'll keep new nodes to things core doesn't already do.
also this one's the simple starter, it's in the name haha, so there's not much to it. the outpaint and inpaint ones are where the nodes actually earn their spot, they do stuff core can't do on its own