Fal API Flux Pro Depth

Depth is generally reckoned to be Flux's strongest ControlNet-style condition - it's the one that reads scene arrangement most reliably out of the whole edges/depth/pose family - and this node is fal.ai's hosted Pro-tier wrapper around Black Forest Labs' own Flux Depth tool, one of the official "Flux Tools" they shipped in November 2024. Unlike the community ControlNet path available on the Dev-tier nodes in this pack, there's no config node to build here: control_image is a plain required input, feeding directly into BFL's dedicated depth-conditioned model.

What it's for. Depth conditioning keeps near things near and far things far - it's the condition to reach for when you want to lock in a scene's spatial layout (a room, a landscape, a figure's pose relative to the camera) while letting the prompt decide everything about what actually appears in that layout. Compared to Canny, it's more forgiving of imprecise source material since it's working with continuous depth rather than hard edge lines.

Inputs. control_image (required) should be a depth map - generate one with any depth estimator (MiDaS, Depth Anything, or whichever depth node you already have in your ComfyUI setup) and feed that map in rather than a raw photo, since depth conditioning specifically wants the estimated depth data, not color information. prompt should read as a full sentence rather than a tag list, matching how Flux's T5 encoder was trained to be used. 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 as strictest-to-most-permissive - a real dial, but one BFL has changed or removed from other endpoints without notice before, so don't build critical logic around it staying exactly as-is. seed is optional.

Output is IMAGE.

What's not here on purpose. No LoRA input - this is the Pro tier, and Pro's weights were never published for anyone to fine-tune, on fal's infrastructure or anywhere else. If depth conditioning plus a custom LoRA is what you actually need, that combination only exists on the Dev-tier ControlNet nodes in this pack, using a ControlNet Config or ControlNet Union Config set to depth mode.

Installing it. Same as every node in the pack: ComfyUI Manager, search "ComfyUI-Fal-API-Flux," install, restart - or clone https://github.com/yhayano-ponotech/ComfyUI-Fal-API-Flux.git by hand (not the placeholder URL the README's manual-install example shows) and pip install -r requirements.txt. Then the non-negotiable step: a config.ini file in the pack's folder with a [falai] section and your api_key.

Troubleshooting. This node has essentially no recorded traffic, so treat any issue as likely generic rather than specific to Depth. If conditioning seems weak or absent, check that control_image is genuinely a depth map and not a raw photo - feeding the wrong kind of image into a depth-specific endpoint is the most common way this class of node produces output that just ignores the structure you intended to lock in. Beyond that: a missing config.ini, an account without Pro-tier access, and the ComfyUI console for fal's real error text.