Say what to change, and FLUX 3 Image Klein edits your picture. A slider shows the picture before and after.
<h3>How to use it</h3>
<ol>
<li>Load your image.</li>
<li>Say the change, then what stays the same: "Change her black jacket to a red leather jacket. Keep her face, her hair, her pose and the street exactly the same."</li>
<li>Run.</li>
</ol>
<h3>What you need</h3>
<ul>
<li>ComfyUI 0.41.0 or newer. That is the first version with FLUX 3.</li>
<li>AusBoss nodes 2.13.0 or newer. In ComfyUI-Manager, search "AusBoss". Already have them? Update first.</li>
<li>The FLUX 3 Image Klein files. They are on Hugging Face behind a license click: open <a target="_blank" href="https://huggingface.co/black-forest-labs/flux-3-image-klein-comfy" rel="ugc">black-forest-labs/flux-3-image-klein-comfy</a>, sign in and accept, then these links work.<ul>
<li><a target="_blank" href="https://huggingface.co/black-forest-labs/flux-3-image-klein-comfy/resolve/main/diffusion_models/flux3_image_klein_sd_int8_convrot.safetensors" rel="ugc">flux3_image_klein_sd_int8_convrot.safetensors</a> → <code>models/diffusion_models</code> · 5.83 GB</li>
<li><a target="_blank" href="https://huggingface.co/black-forest-labs/flux-3-image-klein-comfy/resolve/main/text_encoders/qwen3vl_4b_int8_convrot.safetensors" rel="ugc">qwen3vl_4b_int8_convrot.safetensors</a> → <code>models/text_encoders</code> · 4.86 GB</li>
<li><a target="_blank" href="https://huggingface.co/black-forest-labs/flux-3-image-klein-comfy/resolve/main/vae/flux2-vae.safetensors" rel="ugc">flux2-vae.safetensors</a> → <code>models/vae</code> · 336 MB</li>
</ul>
</li>
</ul>
These are the INT8 files, the same ones ComfyUI's own FLUX 3 template uses. Black Forest Labs says they run on cards with 8 GB. The full-size BF16 files are on the same page for bigger cards (11.1 GB and 8.88 GB): pick them in Diffusion model and Text encoder.
<h3>Tips</h3>
<ul>
<li>Say what stays. End the request with something like "Keep her face, her clothes and the background exactly the same." With that line my test edits moved the picture a median of about 1 px. Without it a big change can do more than you asked: "make it snow" turned a night picture into day.</li>
<li>Ask for the change, don't describe the whole picture. ComfyUI's template suggests a full description of the result. In my tests that redrew the picture, with a new pose and framing, up to 202 px off.</li>
<li>A style change can still move it. A watercolor or another style change can move the picture by tens of pixels, so the slider won't line up exactly.</li>
<li>LoRAs made for FLUX 3 Klein go in LoRA Loader.</li>
<li>About 6 to 8 seconds a run on an RTX 5090.</li>
</ul>
On Qwen Image 2.1 instead: <a target="_blank" href="https://civarchive.com/models/2962616" rel="ugc">Qwen Image 2.1 Edit</a>.
<h3>Credits</h3>
FLUX 3 Image Klein is by Black Forest Labs, under the FLUX Kommunity License: free commercial use for individuals and companies with annual revenue under $5M. Larger companies need a license from Black Forest Labs. The files are Black Forest Labs' own ComfyUI versions.
v1.0 (2026-10-15): first release.
Questions or bugs: <a target="_blank" href="https://github.com/ausboss/ComfyUI-AusBoss/issues" rel="ugc">GitHub issues</a> or the comments here. Videos on <a target="_blank" href="https://www.youtube.com/@ausbossdev" rel="ugc">YouTube @ausbossdev</a>. More from me on <a target="_blank" href="https://x.com/Zanzibased" rel="ugc">X @Zanzibased</a> and <a target="_blank" href="https://github.com/ausboss" rel="ugc">GitHub</a>.

FLUX 3 Image Klein Edit

First release. Needs ComfyUI 0.41.0 or newer and AusBoss nodes 2.13.0 or newer.

Files with SHA256: bb698b6a9cf38117f33d9fa64ff2264031e3cc15a148ca306ecca62eac50f57d