Pay per use
No subscriptions or monthly fees. Only pay for the API calls you make.
Image · Video · Audio models — one API
Every major AI model.
Up to 90% below rival platforms.
We publish our per-request price next to the aggregator you would otherwise buy from, model by model. The gap is checkable, not asserted.
7 models price-checked against rival platforms · as of 24 AUG 2026
Pricing
Up to 90% cheaper than rival platforms
The exact same models you already use, priced against what the major aggregators charge for the same call. Every row is a real, dated price check — compare it yourself.
| Model | Provider | Our price | Rival price | You save |
|---|---|---|---|---|
| Veo 3.1 Fast | $0.015/ SEC | $0.15fal.ai | 90% | |
| GPT Image 2 | openai | $0.0525/ REQUEST | $0.211fal.ai | 75% |
| Nano Banana 2 | $0.04/ REQUEST | $0.08fal.ai | 50% | |
| Nano Banana Pro | $0.075/ REQUEST | $0.15fal.ai | 50% | |
| Nano Banana 2 Lite | $0.0238/ REQUEST | $0.042fal.ai | 43% | |
| Omni Flash | $0.075/ SEC | $0.13fal.ai | 42% | |
| Nano Banana | $0.0312/ REQUEST | $0.039fal.ai | 20% | |
| Gemini Omni 1.1 Flash | $0.09/ SEC | — | — |
→ Rival prices as of 24 AUG 2026
Default configuration · rival list prices
Models
All major AI video models. One integration.
No need to integrate multiple providers. Access everything through one unified infrastructure — at a single per-request price.
Where the price gap comes from
Three structural reasons — not an introductory discount.
01
02
03
generate-video.tsPOST /v1/jobs/submit
const result = await fetch('https://api.e2x.ai/v1/jobs/submit', {
method: 'POST',
headers: {
'Authorization': `Bearer ${API_KEY}`,
'Content-Type': 'application/json',
},
body: JSON.stringify({
model: 'google/veo-3.1-fast-text-to-video',
input: {
prompt: 'walking cat on the clouds',
},
}),
});step 01
One integration
A single request schema covers every image, video and audio model. No per-provider SDK to install, no glue code to keep alive when a provider rewrites its API.
step 02
Cross-provider routing
The same model is served by more than one provider. Each request goes to the one that is cheapest and healthy at that moment, with automatic failover.
step 03
Unit price, no markup tier
You pay the per-request price published on the model page. No subscription, no seat count, no minimum commitment gating the rate you get.
Powered by
One key, every provider we route to
Pricing
Pay only for what you use
Simple, transparent pricing. $100 in = $100 balance. No markups, no surprises.
Transparent pricing
Clear per-request pricing for every model. No hidden costs or surprises.
No lock-in
Start free, scale as you grow. No minimum commitments or contracts.
Per-request price is shown on every model page before you call it
Start building
One key, one request shape, and a smaller bill than rival platforms
Load credits, generate an API key, and send your first job in a few minutes. No subscription, no seat count, no commitment — you only pay for the requests you actually make.
Questions
Every model runs through the same endpoint: POST https://api.e2x.ai/v1/jobs/submit with your key as a bearer token, a model slug, and an input object. The call returns a job id; you poll GET /v1/jobs/{id} until it reports a result. Swapping models means swapping one string — no new SDK, no new auth flow.
Prepaid credits, charged per request. There is no monthly platform fee and no seat pricing — money you load stays as balance until requests consume it. Each model lists its per-request price before you call it, and every charge shows up as a line item in your transaction history.
We route to the same upstream model providers, but we do not stack an aggregator margin on top of an already marked-up price. The comparison table on this page shows our per-request price next to the rival platform we checked it against, with the date of that check — so you can verify the claim model by model instead of taking it on faith.
Failures are not a black box. The job record keeps the upstream error alongside the request that produced it, so a bad input and a provider outage do not look the same in your history. Charges for failed jobs are reviewed and credited back to your balance, and every adjustment appears as its own line in your transaction history.
Yes. Requests are rate limited to keep the platform fair under load; exceeding your limit returns HTTP 429 and you retry with backoff rather than losing the job. Limits vary by account and current platform load — if your workload needs a higher ceiling, get in touch before you build around the default.
Prompts and generated content are not kept permanently after the job completes; temporary caching only exists to re-deliver a result you already paid for. Request metadata and logs are held for a short window for debugging and abuse prevention, then aggregated or deleted. Full detail lives in the privacy policy.
Video, image, and editing models from the providers listed above, all behind one key and one request shape. The catalog is the source of truth — it shows what is live right now, what each model costs per request, and the exact input schema it expects.