FLUX Playground

Opening playground.bfl.ai no longer shows a public generator. The address redirects to a Black Forest Labs sign-in page, and after login the Playground runs inside the BFL dashboard, the same account area that holds API keys, credits and usage reports. The terms of service cover the Playground, including the Dev Playground at dashboard.bfl.ai, under one agreement with a minimum age of 18.

The help center lists the selectable models: FLUX.2 [max], [pro], [flex] and [klein], FLUX.1 Kontext [pro] and [max], and FLUX 3 for video. The FLUX.2 overview page still describes [klein] as API-only, so check the model menu after signing in rather than relying on either page alone. New accounts start with a default organization and project, which matters once credits are shared.

One credit is worth $0.01, and the help center states that Playground and API generations cost the same. FLUX.2 pricing charges the first megapixel of output and then a rate for each further megapixel. For FLUX.2 [pro] that is $0.03 for the first megapixel and $0.015 for each additional one; [klein] 4B starts at $0.014 and [max] at $0.07. Resolution is rounded up to the next whole megapixel.

Editing adds input charges. Each reference image is billed at the reference rate of the chosen model, and when several references are supplied, each is downscaled to and charged as one megapixel. A batch of four costs four times the single price. The practical consequence: an edit with three references can cost more than the output image itself, so decide how many references the task really needs.

BFL positions [klein] for fast, high-volume work, [pro] for production at scale, [flex] for adjustable steps and typography, and [max] for the highest fidelity plus grounding search, where the model can search the web for current products or events. Multi-reference editing accepts up to 10 images in the Playground, against 8 through the API for [pro], [flex] and [max]; [klein] takes up to 4.

Hold the prompt, aspect ratio and output size constant and change only the model. Starting with [pro] gives a reference cost; move to [flex] when lettering matters and to [max] only when [pro] misses detail the brief actually needs. FLUX.1 Kontext is still selectable, but the documentation recommends FLUX.2 for new projects, so avoid building a new workflow around the older model.

The terms say Black Forest Labs claims no ownership of your inputs and outputs, and a help article states that Playground images carry the same commercial usage rights as API images, with no extra licence fee. That permission covers hosted generation. Running open weights such as FLUX.2 [klein] 9B or FLUX.2 [dev] on your own hardware falls under a non-commercial licence unless a separate commercial licence is bought; [klein] 4B is Apache 2.0.

The same terms set limits. You need consent before generating output that depicts a real, identifiable person, may not present output as human-made or as a photograph of a real event, and may not use output to train or fine-tune another AI model. BFL also receives a licence to use inputs and outputs to improve its models; opting out of training means emailing its legal address, and the opt-out does not reach content already used.

Every successful generation has an API Code button that copies its exact parameters as Python, JavaScript or cURL. That is the natural exit once a prompt is stable and you need volume, automation or storage in your own system. API results arrive through polling, and the integration guide warns that generated images expire after 10 minutes, so an application must download each result promptly.

Credits sit at organization level and every member draws from the same pool, while usage is reported per project. Auto top-up can recharge a saved card when the balance falls below a threshold of at least $5, to a target at least $10 higher. Without it, API calls return a 402 error and Playground generation stops when the balance reaches zero.

The quick-start guide suggests adding $10 to $20 of credits to experiment. At FLUX.2 [pro] rates, $10 covers about 333 one-megapixel text-to-image generations, and fewer once edits add reference charges or outputs grow. Editorial assessment: a single designer comparing still images can start near the lower figure, while anyone testing FLUX 3 video should budget separately, because a five-second HD clip already costs $0.85.

Help articles say purchased credits roll over and do not expire while the account exists, but the terms make them non-refundable and non-transferable, including at account closure. Buy in small increments until monthly volume is clear, and switch on auto top-up only after deciding who may approve a card charge.

1. Record the model, output size and starting credit balance.
2. Generate once with the prompt, both references and default settings.
3. Download the image and inspect the label text at full size.
4. Repeat once with FLUX.2 [flex] at the same size and references.
5. Record the balance after each run and inspect the C2PA credential with a verifier.

Accept only a result whose label reads exactly as specified and whose product shape matches the reference. Reject altered text, a changed product or a deduction that differs from the formula. No output or credit figure is reported until the task is run.

For subscription-based image generation, compare Midjourney. Adobe Firefly works inside Adobe apps under its own terms, and Leonardo.Ai offers browser generation with its own plans. Match output size and editing needs before comparing price.