Griptape Driver: Black Forest Labs Image Generation
Flux, over the API, with every dial BFL's service offers
- DRIVER
Most people run Flux locally in ComfyUI. This node is for when you want the same model family but don't want to host it - Griptape Driver: Black Forest Labs Image Generation talks to Black Forest Labs' hosted Flux API, giving you flux-pro, flux-dev, and the newer flux-pro-1.1 / flux-pro-1.1-ultra without a GPU, and billed per generation.
It's also the closest thing this pack has to a "full-featured" image driver, because BFL's API exposes every knob the service offers - and this node exposes all of them. It needs a BFL_API_KEY, which costs money at api.bfl.ai, so the honest framing is: use this if you're building a Griptape agent that generates images on a server you don't control, or if your local box can't run Flux at all.
How it works
Driver → task, like the rest of the pack: the DRIVER output feeds a Griptape Run: Prompt Image Generation Task, which calls the BFL API and returns the image. The node is where you configure the call, and there's a lot to configure.
The inputs worth knowing:
- image_generation_model - seven choices:
flux-pro-1.1(default),flux-pro-1.1-ultra,flux-pro,flux-dev, and the guidance-flavoredflux-pro-1.0-canny,-depth, and-fillvariants. - width / height - dropdowns of values 256–1440 in 32px steps, plus separate aspect_ratio_width / aspect_ratio_height ints. You set either explicit dimensions or an aspect ratio.
- steps and guidance - optional. The tooltips are actually specific here: for
flux-dev/flux-pro, guidance runs 1.5–5 (higher = better prompt adherence, less realism); for the canny/depth variants it's 1–100, where higher means "image over prompt." - prompt_upsampling - BFL rewrites your prompt for more creative output when on. Fun, but it can drift from what you asked for.
- safety_tolerance - strict → none. BFL's safety filter; default medium.
- raw - "less processed, more natural-looking images."
- image_prompt_strength - 0.0–1.0, how strongly a reference image influences the result (relevant for the edit-style models).
- seed - reproducibility.
- api_key_env_var - the env-var name holding your key,
BFL_API_KEYby default. Not the key itself.
Installing
Ships in the ComfyUI Griptape Nodes pack:
- ComfyUI Manager: search "Griptape" → install ComfyUI-Griptape.
- Manual:
cd ComfyUI/custom_nodes && git clone https://github.com/griptape-ai/ComfyUI-Griptape, then restart.
This driver is why the pack's requirements.txt carries a git dependency on griptape-black-forest - if the BFL driver is missing after install, that's the package to check. Manager installs everything automatically. And the torch caveat applies pack-wide (reinstall torch with the CUDA index if Griptape's install breaks ComfyUI's build - see the README).
Gotchas
Two things bite people. First, flux-pro-1.1-ultra and the canny/depth models don't all accept the same parameters - BFL rejects combos that don't apply (e.g., guidance on a model that doesn't take it). If a call errors with a parameter complaint, simplify the node to the defaults first. Second, the raw and prompt_upsampling toggles quietly change output character; if your results don't match your prompt, check those before blaming the model. Cost is the quiet killer too - each generation is a paid API call, and the ultra variants cost more.
Inputs (14)
| Name | Type | Default | Description |
|---|---|---|---|
| image_generation_modelopt | COMBO | flux-pro-1.1 | Select the image generation model. |
| widthopt | COMBO | 1024 | Select the desired image width. |
| heightopt | COMBO | 768 | Select the desired image height. |
| aspect_ratio_widthopt | INT | 169–21 | Select the desired aspect width. |
| aspect_ratio_heightopt | INT | 99–21 | Select the desired aspect height. |
| prompt_upsamplingopt | BOOLEAN | false | Whether to perform upsampling on the prompt. If active, automatically modifies the prompt for more creative generation. |
| safety_toleranceopt | COMBO | medium | Select the safety tolerance level. |
| stepsopt | INT | 0–50 | Number of steps for the image generation process. Values: 1-50 |
| guidanceopt | FLOAT | 0.00 | Optional guidance scale for image generation. For 'flux-dev' and 'flux-pro', values are 1.5-5. High guidance scales improve prompt adherence at the cost of reduced realism. For 'flux-pro-1.0-canny' and 'flux-pro-1.0-depth' values are 1-100 and higher guidance applies image over prompt. |
| intervalopt | INT | 0–4 | Optional interval parameter for guidance control. Values: 1-4 |
| seedopt | INT | 10342349342 | Seed for random number generation. |
| rawopt | BOOLEAN | false | Generate less processed, more natural-looking images |
| image_prompt_strengthopt | FLOAT | 0.100–1 | Image prompt strength. Values: 0.0-1.0. Higher values increase the influence of the image. |
| api_key_env_varopt | STRING | BFL_API_KEY | Enter the environment variable name for the API key, not the actual API key. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| DRIVER | DRIVER | — |