One API for AI Image, Video, and Text Generation

Every image, video, and text model you'd want to call ships its own SDK, its own auth scheme, and its own idea of what a "request" looks like. Aigenway collapses all of that behind one API key and one request shape: provider / model / method / params, the same four fields whether you're generating a still image, a five-second video, or a block of text. Switching models is a string change, not a new integration.

What "one API" actually means here

It doesn't mean one model pretending to do everything — it means one contract in front of many real models from many providers. Every call goes to the same endpoint, POST /v1/generations, with the same envelope. provider and model pick who runs the job, method picks the operation (text_to_image, text_to_video, image_to_video, and so on), and params carries whatever that specific method needs. Only params changes shape between models; the request, the response, and the task states around it never do. List what's live on your account instead of trusting a stale blog post for exact model codes:

curl "https://api.aigenway.com/v1/methods?provider=kling" \
  -H "Authorization: Bearer sk_live_..."

A real call: text-to-image, then swap the model with a string

Here's a text-to-image request. Nothing about the shape below is specific to this one provider — it's the same envelope you'd use for video or text generation:

curl -X POST https://api.aigenway.com/v1/generations \
  -H "Authorization: Bearer sk_live_..." \
  -H "Idempotency-Key: $(uuidgen)" \
  -H "Content-Type: application/json" \
  -d '{
    "provider": "google_banana",
    "model": "imagen-4.0-ultra-generate-001",
    "method": "text_to_image",
    "params": {
      "prompt": "a lighthouse at dusk, storm clouds gathering offshore"
    }
  }'

{
  "task_id": "cmomwuy0m000qbr0371src20a",
  "status": "pending",
  "provider": "google_banana",
  "model": "imagen-4.0-ultra-generate-001",
  "method": "text_to_image",
  "created_at": "2026-09-17T09:12:03.000Z"
}

To try a different image model, or move to video entirely, you change provider, model, and method — the request body's shape, the headers, and how you read the response stay identical. Idempotency-Key makes retries safe: resend the same key after a dropped connection and you get the same task back instead of paying for a second render.

Async by default: one task contract for every method

No generative model returns a finished asset synchronously — a render takes seconds to minutes. Every call above hands back a task, not a result, and every task moves through the same small set of states no matter which provider ran it: pending → processing → succeeded or failed. Poll it:

curl https://api.aigenway.com/v1/tasks/cmomwuy0m000qbr0371src20a \
  -H "Authorization: Bearer sk_live_..."

Or skip the loop and pass a callback_url on the original submit — the gateway pushes a signed POST to your endpoint the instant the task settles, with the same task object you'd get from polling. Either path, the terminal step is identical: read result off the task once status is succeeded, and don't treat result.url as permanent — it's a signed link that expires, not permanent storage. The full state machine, including webhook signature verification, is on the task lifecycle page.

Why one gateway beats a provider account per model

A provider-direct integration works fine for one model. It stops working the moment a second model joins the roadmap — a new SDK, a new auth token, a new billing dashboard, a new request shape to learn, for every provider you add. What one gateway account replaces:

• One key, not N. Authorization: Bearer sk_live_... is the same header for every provider and every method — no per-vendor SDK or credential to rotate.
• One request shape. provider / model / method / params never changes structure; adding a model to your stack is a config change, not a new integration against a new schema.
• One prepaid USD balance. Top up in dollars, spend in dollars, and check what's left with a plain GET /v1/balance — no per-provider credits ledger to reconcile before you can forecast spend.
• One task and webhook contract. Every async job — image, video, or text — exposes the same status values, the same result shape, and the same signed callback format, so one poller and one webhook handler cover every model you'll ever add.

That's the actual case for a single gateway: not that it's "easier," but that the integration work you do once for the first model is the same work that covers the fiftieth.