Transparency report
How credits map to generations, how we moderate content, and how your data is handled. If anything here is unclear, email hello@beingbrush.com.
How beingbrush makes money
You pay a monthly subscription. The only revenue beingbrush earns is that subscription.
We don't sell your data, your prompts, or your images. We don't train on your private work.
What a credit buys
One credit equals one GPT Image 2 generation. Cheaper models cost less than a credit, so the same balance goes further. The rates below are fixed and apply to every tier.
| Model | Provider | Credits / image | Images per 100 credits |
|---|---|---|---|
| GPT Image 2 | openai | 1 | 100 |
| Seedream V4 | fal | 0.6 | 167 |
| FLUX.2 Pro | fal | 1 | 100 |
| FLUX.2 Dev | fal | 0.5 | 200 |
| FLUX.2 Schnell | fal | 0.1 | 1000 |
Weights are derived from each provider's published unit cost and are reviewed quarterly. The full model catalog lives on the pricing page.
The 5% public-image refund
When you publish an image to the public gallery, we refund 5% of that image's credit cost back to your balance, instantly. The refund is recorded in your ledger as a refund entry tied to the generation id.
Content moderation
Every request is screened twice — once on the prompt before generation, once on the image after. We use a dual-model stack so the failure of one provider doesn't leave us exposed.
Primary
nvidia/nemotron-3.5-content-safety
NVIDIA's 4B multimodal guardrail model, served via OpenRouter. Free for low-volume traffic.
Fallback
meta-llama/Llama-Guard-4-12B
Meta's 12B multimodal classifier on Together AI, used when the primary endpoint is rate-limited.
Policy
Content is blocked when the primary or fallback classifier flags it as unsafe across sexual, violence, hate, harassment, self-harm, or illegal-activity categories.
If both moderation providers are unavailable, we fail closed — the request is blocked and no credits are charged. Blocked generations are fully refunded.
Data & image rights
- Private images are yours. They are stored in a private bucket and only ever accessed by you via short-lived signed URLs.
- Public images are displayed in the gallery and licensed under our public gallery terms (see Terms).
- We never use your prompts or generated images to train models.
- You can delete your account and all associated images at any time from settings.
Provider relationships & licensing
beingbrush is a multi-model image generation web application, not an API reseller. We hold commercial accounts with each underlying provider and route generation requests through their official APIs under their standard developer terms. We never expose provider API keys to end users, and we do not offer a programmatic API for others to resell.
| Provider | Models | Output ownership | Commercial use |
|---|---|---|---|
| OpenAI | GPT Image 2 | Customer (you), per OpenAI terms | Permitted |
| fal.ai | Seedream V4, FLUX.2 family | Customer (you), per fal.ai terms | Permitted |
| Replicate | Reserved for future models | Per the model's license on Replicate | Model-dependent |
Each provider's terms take precedence on their respective outputs. beingbrush does not claim ownership of, nor train on, any generated image.
Payments & the merchant of record
All payments are processed by Polar, who acts as the merchant of record — they handle sales tax, invoicing, and payouts to us. Subscription renewals fire a Polar order.paid event that resets your credit balance. Each order id is recorded to guarantee idempotent resets even if Polar retries the webhook.
beingbrush falls under Polar's "AI content generation" category, which is subject to closer review. This page, our Terms, Acceptable Use Policy, and the dual-model moderation stack above are the materials we submit to support that review.