Payment first, then generation
Most AI platforms generate the image and bill you afterwards, or bill you nothing and monetize you some other way. NewEra reverses the order. The 10 NEA payment executes on chain, the backend verifies it against the chain, and only then does generation begin.1
The user approves and pays
A gasless UserOperation calls
generate() on the Generation contract. It charges 10 NEA and emits a payment event.2
The backend verifies the payment
It fetches the transaction receipt, decodes the emitted event directly from its topics, and confirms the payer matches the authenticated smart account.
3
The payment is bound to one image
A transaction hash can be used exactly once. A replayed hash is rejected before generation starts.
4
Generation runs
Only after verification passes does the prompt reach the model.
The verifier decodes the payment event straight from the log topics rather than trusting a stored ABI. Deployed contracts and committed interfaces can drift. Chain data cannot.
The model
Generation currently runs on FLUX.1 [schnell], a fast diffusion model tuned for low step counts, hosted on serverless GPU inference.
Schnell is chosen deliberately. At six steps it produces coherent, detailed images in seconds rather than minutes. For a platform where every generation is a paid on-chain action, latency is not a nice-to-have. A user who waits ninety seconds to find out whether their 10 NEA produced something good will not generate twice.
The model and inference provider are implementation details, not commitments. NewEra may change either to improve quality, latency, or cost. The on-chain cost of 10 NEA per generation is set in the Generation contract.
Provider resilience
A single inference provider is a single point of failure. NewEra runs a provider chain.Primary
The main hosted provider serves nearly all generations.
Fallbacks
If the primary fails or rate limits, the request falls through to secondary providers in order.
Where the image actually lives
This is worth stating plainly, because plenty of projects blur it. The image bytes are not on chain. Storing a megabyte of pixels on BNB Smart Chain would be absurdly expensive and slow, and no serious platform does it. What is on chain is the part that matters: the payment, the payer, and the record that ties the work to your address. The image itself is written to object storage and served from a CDN.On chain
The 10 NEA payment. The payer address. Listings, sales, and the 95/5 split.
Off chain
The image file, served from object storage. The index that makes pages load fast.
The economics of one generation
Every generation is a real cost and a real payment.
The 10 NEA cost is fixed in the Generation contract. It cannot be raised by editing a config file. Changing it would require deploying a new contract, in public, on chain.
Honest limits
AI models are probabilistic. The same prompt does not produce the same image twice, and a prompt that reads well does not always render well.- Output is not guaranteed. NewEra does not warrant that a generated image will match your intent, be original, or be suitable for any particular purpose.
- Models change. Providers deprecate models. Quality characteristics shift. We will change models when a better one exists.
- Generation can fail. Providers rate limit and go down. When every fallback fails, the generation fails.
How NEA moves through the platform
Welcome credits, generation cost, sale splits, and rewards.
NEA is a utility credit for using NewEra. It is not an investment, and it cannot be withdrawn or exchanged for money outside the platform.
