ComfyUI-Floyo-Flux2-API-node Introduction
ComfyUI-Floyo-Flux2-API-node brings Black Forest Labs’ FLUX.2 [pro] and [flex] image generation and editing into a ComfyUI workflow. You can create an image from a written prompt or edit an existing image using a prompt and reference images.
The extension is designed to work with Floyo’s URL-based storage pipeline. It sends a request to the FLUX.2 API, waits for the result, and handles the returned image so it can continue through your workflow. This lets you use an online image model from ComfyUI without managing the model locally.
How ComfyUI-Floyo-Flux2-API-node Works
Think of a node as one step in a creative production line:
- Describe the result. Enter a prompt, such as “a ceramic teapot on a sunlit kitchen table, editorial photography.”
- Choose a starting point. For text-to-image, the prompt is enough. For an edit, connect an existing image and describe what you want changed. You can add more reference images where the node supports them.
- Choose your controls. Set the image dimensions and, if useful, a seed, output format, or other model-specific settings.
- Run the workflow. The node sends the request to the API and checks its status while the image is being made.
- Use the result. The listed ComfyUI nodes provide an
IMAGEoutput that you can connect to a Save Image node or another compatible image node.
The README also describes a signed output URL used in Floyo’s storage pipeline. That URL expires quickly, so a workflow that uses it should pass it to the next Floyo download or output node promptly. The documented node outputs below, however, are IMAGE tensors. In practical terms, an IMAGE connection carries an image within ComfyUI; a URL is an address from which an image can be downloaded. Check the socket type on the node you are using before connecting the next step.
ComfyUI-Floyo-Flux2-API-node Features
Text-to-image generation
The text-to-image nodes turn a prompt into a new image. Set width and height to choose its shape and size. For example, a wider canvas suits a landscape scene, while a taller one may suit a portrait. The documented dimensions must be multiples of 16.
Image editing and references
The image-edit nodes begin with a required input_image. Your prompt describes the desired edit, and optional additional images can serve as references. For example, you might provide a product photo as the main image and another image as a visual reference for the setting.
Keep within the limit for your chosen model: up to 8 total input images for [pro], or 10 total input images for [flex]. The original README also specifies a 9-megapixel total reference-image limit for [pro].
The nodes accept ComfyUI IMAGE inputs and convert them for the API. The original README also mentions URL fields for supplying images by URL, although it does not list those fields in the node-by-node input descriptions.
Creative and output controls
- Seed: Set a specific seed when you want a consistent starting point for comparisons. A value of
-1selects a random seed. - Safety tolerance: The documented range is
0–6. Use the setting appropriate to your project; the README does not define what each level permits. - Output format: Choose
jpegorpngaccording to the format you want for the generated image. - Guidance and steps: These are available on the [flex] nodes. They give you additional controls to experiment with, though the original README does not specify a guaranteed visual effect for any particular value.
The extension also polls the API while a request is pending and includes basic error handling and timeouts. You do not need to manually check whether a generation request has finished.
ComfyUI-Floyo-Flux2-API-node Models
The extension documents two FLUX.2 model options. Each has a text-to-image node and an image-edit node.
| Model and node | Use it when | Documented inputs and output |
|---|---|---|
FLUX.2 [pro] Text-to-Image (Flux2ProTextToImage) | You want to make a new image from a prompt. | Requires prompt, width, and height. Offers optional seed, safety_tolerance, and output_format. Outputs an IMAGE for a node such as Save Image. |
FLUX.2 [pro] Image Edit (Flux2ProImageEdit) | You want to change an existing image, optionally using other images as references. | Requires prompt and input_image. Accepts input_image_2 through input_image_8 and optional size, seed, safety, and format controls. Outputs an IMAGE. |
FLUX.2 [flex] Text-to-Image (Flux2FlexTextToImage) | You want to generate an image with additional guidance and step controls. | Requires prompt, width, height, guidance (1.5–10), and steps (1–50). Also offers optional seed, safety, and format controls. Outputs an IMAGE. |
FLUX.2 [flex] Image Edit (Flux2FlexImageEdit) | You want to edit an image with more reference-image slots and [flex] controls. | Requires prompt and input_image. Accepts input_image_2 through input_image_10, plus optional size, seed, safety, format, guidance, and steps controls. Outputs an IMAGE. |
For [pro] edits, setting width or height to 0 uses the input image’s dimension for that setting. The README specifies a 4-megapixel maximum output resolution and gives 2048 × 2048 as an example. It does not provide a model quality or speed comparison, so choose based on the controls and reference-image capacity you need.
Troubleshooting ComfyUI-Floyo-Flux2-API-node
The node reports an API error or does not produce an image
- Read the error shown in the node output or ComfyUI’s execution details. The node handles pending, ready, error, and moderation states.
- Confirm that the API key used by the extension is valid. The documented configuration options are
config.iniand theBFL_API_KEYenvironment variable. - Check that your prompt and images are appropriate for the API’s safety requirements.
- Try a simpler request with fewer reference images. If the request times out, run it again after checking that the API service is reachable.
An edit fails when several references are connected
- Count the main
input_imageas one reference. - For [pro], use no more than 8 images in total and observe its documented 9-megapixel total reference limit.
- For [flex], use no more than 10 images in total.
- Remove extra references and run the workflow again.
The requested dimensions cause an error
- Check that nonzero width and height values are multiples of 16.
- Keep the requested output within the documented 4-megapixel maximum.
- For a [pro] image edit, try
0for a dimension when you want it to match the corresponding dimension of the input image.
A downstream Floyo node cannot retrieve the result
A signed URL can expire quickly. Run a workflow that downloads from that URL promptly after generation. Also check the connection type: the four nodes listed in this README document an IMAGE output, so a downstream node expecting a URL may need a different connection or workflow step.
Frequently asked questions
Do I need a starting image?
Only for an Image Edit node. Text-to-Image starts from your prompt.
Can I use the same seed for another attempt?
Yes. Enter a specific seed rather than -1 when comparing attempts. Other changes to your settings or inputs can still change the result.
Which model should I choose for more reference images?
The documented limit is 10 total images for [flex] editing and 8 for [pro] editing.
Can I connect the result to Save Image?
Yes. Each node’s documented output is an IMAGE, which is suitable for a compatible image node such as Save Image.
Learn More about ComfyUI-Floyo-Flux2-API-node
- Black Forest Labs documentation (https://docs.bfl.ai/quick_start/introduction) provides the official starting point for learning about the FLUX API and its models. FLUX.2 documentation remains available there. (docs.bfl.ai (https://docs.bfl.ai/quick_start/introduction))
- Black Forest Labs prompting guide (https://docs.bfl.ai/guides/prompting) can help you explore how to write prompts for image creation and editing. (docs.bfl.ai (https://docs.bfl.ai/quick_start/introduction))
