Hey folks, hoping someone can spot what I'm missing here. I'm integrating Playground's image generation API into a user feedback dashboard (showing "concepts" based on feature requests, it's a fun internal tool). My issue is that the images coming back are consistently low resolution, even when I'm explicitly setting the `hd` parameter.
I'm using the `image-generations` endpoint with the `playground-v2.5-1024px-aesthetic` model. According to the docs, this model should support 1024x1024, and I'm passing `"hd": true` in the payload. The images *look* okay, but they're coming through at 512x512 when I check the dimensions. It's causing them to appear blurry when displayed full-width in our dashboard.
Here's a simplified version of my call:
```json
{
"model": "playground-v2.5-1024px-aesthetic",
"prompt": "a sleek dashboard graph showing rising user engagement",
"aspect_ratio": "1:1",
"hd": true,
"steps": 50
}
```
I've tried both `"hd": true` and `"hd": false`, and the output size doesn't seem to change. I'm also not hitting any rate limits or getting errors. Is the `hd` parameter for something else, like an internal upscaling step, and I need to specify the size differently? Or is there another model I should be using for true 1024px output?
From a product analytics angle, this is a bit of a blocker—the low-res images make the whole feature look unpolished. Any insight would be super appreciated!
Interesting. I've used that same model but hadn't hit this. My first thought: is the `hd` param actually supported by that specific model? The docs sometimes lump parameters together, but they can be model-dependent.
Could you check the response headers for the image's actual content-length or metadata? Sometimes the image gets resized post-generation on their end before delivery, which is super annoying. I've also seen weirdness where `aspect_ratio` interacts with the size output.
If you're on Python, can you share how you're extracting and checking the dimensions? Just in case it's a PIL or OpenCV read issue on your side.
Data is the new oil - but it's usually crude.
I've been testing the API for a cost analysis project and ran into a similar size discrepancy. Could the pricing be different for the actual 1024px output versus the 512px you're receiving? That would be a big factor in the total cost of ownership.
Do the API usage logs or billing line items show any detail on resolution tiers? If they're charging for HD but delivering SD, that's a billing issue as well as a technical one.