Available through FAL — FLUX

flux-dev

flux-dev on fal_ai · Fal.ai

What it takes and gives

a prompt → an image · seed, strength, steps, guidance, aspect ratio, variations

Best fordeclared, not tested

General-purpose concept art and character portraits.

Good at

prompt adherencecoherent anatomypainterlyphotoreal

Not for

no negative prompt — it is guidance-distilled

Licence

FLUX.1 [dev] Non-Commercial License v2.0 · Commercial use: under Fal.ai’s terms

Made by Black Forest Labs

Every model's licence

Price

About 3.25 credits per image · measured on real runs

What comes out

Production quality — 1024px typical on the shorter side, across 12 images

Speed and reliability

5 runs in 30 days · typically 4.6 s · Fal.ai: 9 runs in 30 days

Also runs on

flux-schnell on Fal.ai · ≈ 0.39 credits · −2.86In app · own key on API
flux-2-klein-4b on Fal.ai · ≈ 0.91 credits · −2.34In app · own key on API
flux-pro on Fal.ai · ≈ 3.25 credits · sameIn app · own key on API
qwen-image on Fal.ai · ≈ 3.25 credits · sameIn app · own key on API
sd-3.5 on Fal.ai · ≈ 3.25 credits · sameIn app · own key on API
sdxl on Fal.ai · ≈ 3.25 credits · sameIn app · own key on API

Run it from code

With a team API key and a project id. The run answers with a jobId; poll GET /api/v1/runs/{jobId} for the image.

# 1. Once: a workflow that runs flux-dev (returns workflowId)
curl -X POST "https://app.sushibelt.ai/api/v1/workflows" \
  -H "Authorization: Bearer $SUSHIBELT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"name":"flux-dev","projectId":"'"$PROJECT_ID"'","steps":[{"key":"subject","type":"TextEntry","config":{"text":"a red fox in the snow","apiInputKey":"prompt"}},{"key":"image","type":"DiffusionWorker","config":{"providerKey":"fal_ai","model":"flux-dev"},"inputs":["subject"]}]}'

# 2. Run it, with your own prompt
curl -X POST "https://app.sushibelt.ai/api/v1/workflows/$WORKFLOW_ID/runs" \
  -H "Authorization: Bearer $SUSHIBELT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"inputs":{"prompt":"a red fox in the snow"}}'

Where it runs

In app · own key on APIRuns in the app on Sushibelt's Fal.ai account. Through the API it runs only on your own Fal.ai key.

Runs on your own Fal.ai key, or on credits

On your own key, Fal.ai’s terms apply to your team directly.

What we learned

img2img keeps the reference's composition at every strength — a full-body reference will not become a close crop, however the prompt is worded. It does re-pose at about 0.9: a front-facing reference gives a real back view. For a different framing, generate from the brief with the same seed and style rather than from the image.

How it’s used →