My results after the trial: Pro isn't worth it if you only need 10 images a week.

I run a three-person team building internal dev tools, and we've been piping image generation into our CI pipeline for auto-generating release mockups and feature screenshots. We're on GitHub Actions, calling the API directly from a shell step. Our weekly output is around 8-15 images depending on that week's deploys, so we landed right on the fence between free and Pro.

Here's what I've seen after running both tiers in production for about two months:

- **Weekly quota vs burst need** - Free gives you 10 images per billing week, no rollover. We hit that cap twice when we had a big feature release and needed 20 images in one day. Pro bumps it to 50 or 60 (the $16 plan gives roughly 50, not 40 extra as OP says -- check your dashboard, it's 50). If you ever run a batch job (say, regenerating all your product screenshots after a UI refresh), the free cap forces you to spread it over two weeks.
- **Latency difference** - We timed it with a simple `curl` wrapper. Free tier averaged 2.5 seconds per image on the first call of the day, dropping to ~2.0 on warm cache. Pro averaged 1.2-1.4 seconds consistently, including cold starts. For a single image that's negligible, but when your CI pipeline has to generate 10 images sequentially, that's an extra 10-13 seconds of wall-clock time. For us, that meant a 30-second CI job turned into 45 seconds, which started to annoy the team during short merges.
- **Watermark removal** - OP says free watermarks are barely visible, and they are -- it's a small logo in the bottom-right corner at ~10% opacity. But we used the images in public changelogs and on a marketing site, and the watermark had to go. We tried post-processing with ImageMagick to blur it, but it looked sloppy. Pro removes it entirely. If you're shipping these images to customers, that's a hard requirement.
- **Rate limits and error handling** - Free tier rate-limits at 1 request per 3 seconds. Our script originally did rapid-fire calls and hit 429 errors. We had to add exponential backoff (simple 2s sleep). Pro allows up to 3 concurrent requests, so we could run batches faster without handling retries. The retry logic itself added maybe 50 lines of YAML and a bash loop, which is fine, but it's a gotcha if you're integrating quickly.
- **Support responsiveness** - I filed a ticket on the free plan about a recurring 500 error during peak hours (weekday afternoons). Got a canned reply three days later pointing at their status page. On Pro I filed a similar ticket about an authentication token bug -- got a human response in 4 hours with a workaround. For a hobby project that's fine, but if your CI pipeline failing blocks deployment, that turnaround matters.

My pick: If you can confidently stay under 10 images per week *and* never need to burst beyond that *and* you're okay with the tiny watermark in your outputs, save your $16 and stay on free. But the moment you hit that cap once, or you need clean images for customer-facing content, Pro pays for itself in headache avoided.

What's your deployment setup? Are you calling the API from a CI runner or manually? If it's manual, your original take is solid. If it's automated, tell us how many concurrent jobs you usually run -- that'll tip the scale.