Fal API Flux Pro

If Fal API Flux Dev is the one everyone reaches for, Fal API Flux Pro is the one you reach for when Dev's output is annoying you. FLUX.1 [pro] is Black Forest Labs' undistilled base model - the version Dev and Schnell are both compressed down from - and it's never been released as downloadable weights, only ever sold through an API. This node is fal.ai's wrapper around that API, so it's the closest a ComfyUI graph gets to the real thing rather than the distilled copy.

The practical reason to care: Dev's guidance distillation is the same process blamed for Flux's infamous homogenized faces (the "everyone looks like an alien" complaint that follows Flux around). Pro skips that distillation step, so it's noticeably better at facial variety and doesn't need the guidance-scale workarounds Dev users trade tips about. You pay for that in dollars per call rather than in local GPU time, which is the whole trade this node exists to make.

Inputs that matter. Same core shape as the Dev node - prompt (multiline, write it as a sentence), width/height (1024 default, step 8), num_inference_steps (28 default, 1–100), guidance_scale (3.5 default, 0.1–40), num_images (1–4), and enable_safety_checker. What Pro adds on top is safety_tolerance, a six-step enum (1–6) that fal documents running from strictest to most permissive filtering. It's worth knowing that dial exists rather than fighting a rejected generation blind - and worth knowing it isn't bulletproof: Black Forest Labs has pulled this same parameter out from under users before, silently removing safety_tolerance from their Kontext Pro/Max API in mid-2025 with no changelog and no warning. If the field stops doing anything one day, that's not a bug in this pack, that's BFL changing their API underneath it. seed is optional, defaults to 0 for a random draw.

Output is a single IMAGE.

What's missing on purpose. There's no LoRA input here, and there isn't going to be - Pro's weights aren't published, so nobody can fine-tune it. If your workflow depends on a specific LoRA, that's the strongest argument for staying on one of the Dev-with-LoRA nodes instead of upgrading to Pro.

Install and setup are identical to every other node in this pack: install via ComfyUI Manager (search "ComfyUI-Fal-API-Flux") or clone the repo manually into custom_nodes and pip install -r requirements.txt, then drop a config.ini in the pack's folder with your fal.ai key under [falai]. Restart ComfyUI after either path.

Troubleshooting. Because Pro is the priciest tier fal offers here, an auth or account-tier problem is more likely to show up as a confusing billing-adjacent error than a plain "invalid key" message - check that your fal.ai account specifically has Pro-tier FLUX access, not just Dev, since some accounts are provisioned more narrowly than others. Beyond that it's the same short list as the rest of the pack: bad or missing config.ini, and reading the actual ComfyUI console output rather than trusting the node's own error text, since that's where fal's real response lands.