Image Upscaling API: Pricing Compared (2026)
Per-image pricing for the main image-upscaling APIs, using each provider's own published rates — plus what actually changes if you're processing 10 images a month versus 10,000.
🎯 Quick Answer
If you just want low-volume, pay-as-you-go upscaling with no monthly minimum, BigImg's API comes out cheapest per image (~$0.06) and has a genuine free tier (10/month, no card). If you're processing tens of thousands of images a month, calling Replicate directly is cheaper in raw compute — but you're building and hosting the auth/queue/retry layer yourself.
Pricing below reflects each provider's own published rates as of September 2026 — API pricing changes often, so check the provider's current page before committing to a plan.
Side by Side
| Provider | BigImg | Let's Enhance | DeepAI | Replicate (DIY) |
|---|---|---|---|---|
| Free tier | 10 images/month, no signup | 5 images/month | Small trial credit only | None — pay per call from image one |
| Monthly minimum | None | $12/mo for the 100-image tier | ~$9.99/mo for the Pro credit tier | None |
| Cost per image beyond free | ~$0.06 (credit pack) or unmetered on $10/mo Unlimited | ~$0.12 at the 100-image tier | 1–6 credits/image depending on settings | ~$0.003–0.005 in raw model compute |
| Setup | 1 key from your dashboard | API key | API key | You build auth, queueing, retries, storage |
| Max scale | 16x | 4x | 4x | Depends on chosen model |
Where the cheap DIY number is misleading
Real-ESRGAN itself is genuinely cheap to run — under half a cent per image in raw GPU compute on a service like Replicate. That number gets quoted a lot as "why pay for an API", but it only covers the model call. It doesn't include the auth layer, rate limiting, retry logic for failed predictions, file size/format validation, or somewhere to host the output so it doesn't just live on a 1-hour-expiring delivery URL. For a side project calling it a few times a week, none of that matters. For anything customer-facing, it's a few days of work you're not billing anyone for.
When a subscription-based API wins
Let's Enhance and DeepAI both make more sense than pay-per-image once you're consistently past their free tier and running steady volume every month — a fixed monthly fee against predictable usage is easier to budget than metered credits, and both offer more editing features beyond plain upscaling (denoise presets, face enhancement, batch tooling) if you need those in one call.
When BigImg's API is the better fit
- You're testing an idea. 10 free images a month with no card is enough to prototype an integration before spending anything.
- Your volume is spiky, not steady. Credits never expire, so a quiet month doesn't cost you a subscription you didn't use.
- You already use BigImg's web tools. API calls and website usage draw from the exact same balance — no second account, no separate top-up.
- You need higher scale factors. 8x and 16x aren't offered by either of the other two providers compared here.
Try it with your first 10 images free
Create a key from your dashboard, no card required.
Read the API docs →