flux_3_api

Flux 3 API — ComfyUI Nodes

ComfyUI nodes for the BFL Flux 3 image and video API, plus an LLM-powered prompt generator.

Nodes

• Flux 3 Image (API) — generates or edits images via POST https://api.bfl.ai/v1/flux-3-image (Docs, API reference).
• Flux 3 Video (API) — generates video via POST https://api.bfl.ai/v1/flux-3-video (Docs, API reference).
• Flux 3 Video Upscale (API) — super-resolution for short clips (≤20s, ≤50MB) via POST https://api.bfl.ai/v1/flux-tools/video-upscale-v1 (Docs, API reference). Precise mode preserves identity; creative mode enhances detail. Output capped at ~14.4 MP per frame; source audio preserved.
• Flux 3 Openrouter Prompt — turns a vague idea into a structured FLUX 3 prompt using an OpenRouter LLM, guided by a prompting skill. Output feeds straight into the Video node's prompt input.

Setup

Add your keys to .env:

BFL_API_KEY=bfl_...
OPENROUTER_API_KEY=sk-or-...

Get an OpenRouter key at https://openrouter.ai/keys. Keys are looked up in this order: node api_key field → .env → environment variable.

Optional in .env:

BFL_BASE_URL=https://api.bfl.ai
BFL_IMAGE_BASE_URL=https://api.bfl.ai

(BFL_BASE_URL is the host for video and upscale, BFL_IMAGE_BASE_URL the host for the image node. Both default to https://api.bfl.ai; only change them if BFL announces a different host.)

---

Flux 3 Image (API)

Text to image, or edit / restyle / combine 1–10 reference images on the images input (each 256×256 to 16 MP). There is no mode: describe in the prompt how the references are used ("Turn Image 1 in the style of Image 2").

Parameter	Values
aspect_ratio	auto (first reference image, else 1:1), 21:9 … 9:21
resolution	768sq, 1k (default), 2k, 4k (can take several minutes)
safety_tolerance	0–4 (default 2), 0 = strictest
grounding	web and image search before generating (default on)

The API has no seed; the seed widget only makes the node run again. For layouts with bounding boxes, name elements as <id> in the prompt and end it with a JSON list of rows, see Layout and editing.

Outputs: image and metadata (including cost and output_mp).

---

Flux 3 Video (API)

Outputs: video (mp4, with audio track) and metadata. Feed straight into Save Video.

Modes (mode)

One mode per request; the rest of the request stays the same.

mode	requires	does
t2v	–	Text → Video
i2v	images (1–10)	Image(s) → Video (keyframes)
v2v	video	Continue a clip (start_video)
draft_enhance	draft_cache	Final-render a previous draft run

Parameters

Field	Required	Values
mode	always	t2v i2v v2v draft_enhance
prompt	t2v/i2v/v2v	free-form
keyframes	i2v	1 image (opening frame) · 2 images (start+end) · 3–10 images (evenly spread, needs duration) · or [seconds, image] pairs (storyboard). URL or base64.
start_video	v2v	mp4 URL or base64
draft_cache	draft_enhance	base64 bundle or URL from the draft output of a previous run
aspect_ratio	–	auto (default), 21:9, 2:1, 16:9, 4:3, 1:1, 3:4, 9:16
duration	–	whole seconds 5–20 or auto (default)
resolution	–	hd (default) or fhd
generate_audio	–	bool, default true
safety_tolerance	–	0 (strictest) to 4, default 2. Capped at 2 when conditioning media is present.
draft	–	bool. true = fast hd preview; result in