Fal API Flux Dev Canny With LoRA

Don't confuse this with the generic "with LoRA and ControlNet" node elsewhere in the pack - that one needs you to build a separate ControlNet Config node and wire it in. This one skips all of that: control_image is a plain required input right on the node itself, no config node in between. That schema shape strongly suggests it's calling Black Forest Labs' own Flux.1 Canny [dev] tool directly (one of the official "Flux Tools" BFL shipped alongside Fill, Depth, and Redux back in late 2024) rather than a general community ControlNet, which is also why it's Canny specifically and not a configurable condition type.

Inputs that matter. control_image is required, and here's the honest gap: the schema doesn't say whether it wants a raw photo or an already edge-detected map. Official Canny-conditioned Flux tools commonly run edge detection server-side on whatever you hand them, so feeding a plain photo is the reasonable first attempt - but if the output looks like it's ignoring your structure entirely, running an actual Canny preprocessor node first and feeding that map in instead is the next thing to try. prompt should read like a sentence, not a tag list, since Dev's T5 encoder is built for natural language. width/height default to 1024×1024 in steps of 8, num_inference_steps defaults to 28 (1–100), guidance_scale defaults to 3.5 (0.1–40), num_images batches 1–4, and enable_safety_checker toggles the content filter. seed is optional. For LoRAs, lora_1 through lora_5 each take a LORA_CONFIG from a "Fal API Flux LoRA Config" node - stack up to five, none required.

Output is IMAGE.

Where this fits versus the alternatives. If you want edge-guided generation with a LoRA and don't need to also configure a separate ControlNet with a chosen path or conditioning strength, this node is simpler to wire than the generic combo node. If you need depth or pose instead of edges, or you want to fine-tune conditioning strength directly, you'll want the ControlNet Config or ControlNet Union Config nodes feeding into the generic combo node instead - this one has no conditioning_scale of its own.

Installing it. Standard for the pack: ComfyUI Manager, search "ComfyUI-Fal-API-Flux," or clone https://github.com/yhayano-ponotech/ComfyUI-Fal-API-Flux.git manually (the README's own manual-install example points at a placeholder repo, not this one) and pip install -r requirements.txt. Then a config.ini with a [falai] section and api_key in the pack's folder - no exceptions, no in-UI alternative.

Troubleshooting. This node barely registers any traffic, which tracks with it being a specific, less-discovered corner of an already obscure pack - so if edges seem ignored, work through the basics before assuming it's broken: confirm control_image is actually connected (not just present but wired to something meaningful), try both a raw photo and a pre-processed edge map since the node doesn't document which it expects, and check that your LoRA URL resolves to a raw .safetensors file rather than a page URL, per the README's own troubleshooting advice.