Fal API Flux Pro Canny

Flux Canny is one of the four official "Flux Tools" Black Forest Labs shipped in November 2024 (alongside Fill, Depth, and Redux), and this node is fal.ai's hosted Pro-tier version of it - edge-guided generation, called over the network, no local ControlNet checkpoint or config node required. control_image is just a plain required input on this node; there's no separate ControlNet Config to build and wire in, because BFL's Canny tool isn't a general-purpose ControlNet, it's a dedicated model trained specifically for this one condition.

That's the practical upside of picking this node over the Dev-tier ControlNet path: fewer moving parts, one image in, one image out. The tradeoff is the usual Pro-tier one - no LoRA support, since Pro's weights are API-only and nobody outside BFL can fine-tune them, and you're paying fal's Pro-tier rate per call instead of Dev's.

Inputs. control_image (required) is your structural reference - feed it a photo and let fal's Canny pipeline extract edges, or run an explicit Canny preprocessor first if results feel like they're ignoring your structure; the schema doesn't document which behavior to expect, so testing both quickly settles it. prompt should read as a full sentence - Flux's T5 encoder is built for natural language, not tag soup. 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, enable_safety_checker toggles the content filter, and safety_tolerance is a 1–6 enum fal documents running from strictest to most permissive - worth knowing it exists if a generation comes back unexpectedly filtered, and worth knowing BFL has removed this exact parameter from other endpoints without warning before, so treat it as informational rather than guaranteed to stick around forever. seed is optional.

Output is IMAGE.

How this compares to the Dev-side Canny options. This pack also has a Dev-tier "Canny With LoRA" node and a fully generic ControlNet Config path for Dev. Reach for this Pro node when you want Canny conditioning without any LoRA and don't mind paying Pro rates for Pro's stronger prompt adherence and lack of the same-face collapse Dev is known for; reach for the Dev nodes when you need a custom LoRA in the mix, since Pro simply doesn't support that.

Installing it. No node-specific steps: ComfyUI Manager, search "ComfyUI-Fal-API-Flux," or a manual clone of https://github.com/yhayano-ponotech/ComfyUI-Fal-API-Flux.git (the README's own git-clone example has a placeholder owner in the URL, so grab the real one) plus pip install -r requirements.txt. Then a config.ini with [falai] and api_key in the pack's folder - this node, like every other one here, does nothing without it.

Troubleshooting. With near-zero measured usage on comfy.icu, this is a rarely-touched corner of an already obscure pack, so lean on the pack-wide basics first: confirm config.ini is actually present and correctly formatted, that your fal account has Pro-tier Flux access specifically (not just Dev), and read the ComfyUI console rather than trusting the node's own error surface for what actually went wrong.