Video wird geladen...
Video konnte nicht geladen werden
Qwen-Image-2.1 is now supported in ComfyUI! Open weights. One 7B checkpoint that generates and edits. → Image generation at native 2K → Instruction editing from up to 10 reference images in a single pass → RGBA output, alpha included
168,022 Aufrufe • vor 1 Tag •via X (Twitter)
26 Kommentare

Try the workflows!

Where are the custom nodes at? No update for the desktop app yet & node errors within the app.

测评

having native 2K generation and RGBA alpha channel output working directly inside ComfyUI on day zero is fantastic. tested 4-bit quant and it fits comfortably under 5 GB VRAM.

@grok what is the license? Free for commercial use?

7B, native 2K, multi-reference editing, and RGBA output in one open-weight model. 🔥 ComfyUI just made serious image workflows a lot more flexible.

is there a prompt guide for this model?

open weights in comfyui means designers finally have a local option that doesn't phone home to some api. 7b model doing 2k native with rgba is actually useful — no more exporting, uploading, waiting for responses. referencing 10 images in one pass though — that's the real move for product design workflows. do a quick edit without round-tripping through api calls

ComfyUi catching up too fast. Great for us users.

Native 2K plus ten reference images in one pass is a workflow, I can test it locally.

Liking it so far

Very cool 😎

The 10-reference edit path is the standout: it could cut asset-prep work. The production test is consistency across iterative edits—does the subject, layout, and transparency survive several revisions? That’s what determines whether native 2K + RGBA saves time beyond a demo.

El salto no es solo generar a 2K: combinar referencias, edición por instrucciones y salida RGBA acerca estos modelos a flujos reproducibles. En contaduría y auditoría, podría ayudar a documentar evidencia visual y crear materiales consistentes, siempre con fuentes, trazabilidad y revisión humana.

covered this in today's daily digest – open-weight unified generation and editing at 7b, already outperforming most closed-source models. fast support from comfyui makes self-hosting even easier

Native 2K editing on open weights is the part that actually matters when product shots cannot leave the building. Cloud still covers the rest; this is for the cases where it cannot.

I am starting on the edit workflow not the t2i one.

Great insights! AI is rapidly transforming how we approach productivity and innovation. Excited to see where this technology leads next.

I wish there was something in Comfy UI that can detect models with workflows and help me delete models with workflows. I am running out of space and cleaning up comfy UI is a pain.

Having been challenged by the question "what are the benefits of AI" I initially reacted with simple things, but really, it's like the old question "what are the benefits of the internet"

That’s a big update. 2K gen, 10-image editing, and RGBA in one 7B open checkpoint is wild.

Goooooooood morning ✌️

RGBA output and reference-conditioned edits are the workflow-changing features here. The real test is whether alpha edges survive repeated edits and whether the same graph stays reproducible across local and cloud runs.

Amazing

Huge one 7B open checkpoint doing native 2K generation + multi-image editing with RGBA is a game changer for ComfyUI workflows!

Lets make some Q4 or Q5, Q6
