FluxKontextImageScale
When your image isn’t the right size, and you don’t want to argue with it. This node gets it resized properly so your workflow doesn’t explode later.
🧠 What This Node Does
The FluxKontextImageScale node is a specialized image resizing tool from the comfy_extras.nodes_flux module. It scales your image in a context-aware, no-nonsense way — meaning it’s perfect for workflows where size actually matters (looking at you, conditioning and machine learning models).
It’s especially useful in complex setups that need consistency, like multimodal processing (image + text + pose, etc.), where the slightest dimensional mismatch can trigger the kind of error that makes you question your life choices.
🧩 Node Type
- Category:
Advanced → Conditioning → Flux - Python Module:
comfy_extras.nodes_flux - Output is List: ❌ No
- Output Type: Image tensor
- Resizing Type: Context-aware, no quality-enhancing filters — just straight-up dimensional adjustment
🔌 Inputs
| Name | Type | Required | Description |
|---|---|---|---|
image | Image | ✅ | The image to resize. Must be a valid tensor. ComfyUI-compatible formats only (.png, .jpg, etc.). |
🔁 Outputs
| Name | Type | Description |
|---|---|---|
IMAGE | Image | The resized image. Ready for conditioning, model input, or multi-modal fusion. |
⚙️ Parameters
This node has no user-facing parameters — it just does the job based on the context of your pipeline and where it’s placed.
Internally, it performs smart resizing that aligns the image dimensions with what the rest of your model expects, especially in workflows built around Flux Kontext structures.
Basically: no knobs to twist — just plug it in, and it sizes things properly behind the scenes.
✅ Recommended Use Cases
- Pre-conditioning cleanup
Align image sizes before sending to ControlNet, CLIP, or LoRA inputs. Because mismatched resolutions make them sad. - Multimodal workflows
Scaling images to match other forms of input (pose, depth, text, segmentation maps, etc.). - Cloud/Remote Workflows
Slim down oversized images before sending them across the internet (save your bandwidth and your patience). - Preprocessing for ML pipelines
Normalize image dimensions before inference or training. No model likes a surprise.
🧪 Workflow Setup Example
[Load Image] ↓ [FluxKontextImageScale] ↓ [CLIP Text Encode / Conditioning Node / ControlNet Preprocessor] ↓ [KSampler or Diffusion Node]
This node should go before any node that depends on fixed or known image dimensions. Stick it early so the rest of your pipeline doesn’t break when it expects a 512x512 and you gave it a 498x505.
💬 Prompting Tips
- You don’t prompt this node directly. But if your prompt depends on alignment between the image and the conditioning (like pose or depth), this node makes sure your sizes are aligned behind the scenes.
- Great for image-to-image prompting where you need to control resolution before it hits the denoising loop.
🔥 What-Not-To-Do-Unless-You-Want-a-Fire
- Don’t feed this node corrupt or invalid image tensors. It will break, and it will blame you.
- Don’t expect it to enhance image quality — this is not an upscaler. It scales, it doesn’t beautify.
- Don’t assume this will magically fix bad dimensions later in the pipeline. Use it early.
- Don’t skip it if you’re using multiple conditioning inputs that must be aligned — you will get shape mismatch errors.
- Don’t expect output size control via UI — this node handles things behind the scenes for you.
🧾 Final Thoughts
FluxKontextImageScale is the quiet hero in your ComfyUI setup. No fancy sliders, no noise, just reliable, context-aware resizing to keep your workflow from falling apart due to pixel drama.
If your pipeline depends on precision (and let’s face it, they all do), this node belongs in your toolkit.