SDXL Turbo on QuickSilver Pro
SDXL Turbo is an open, single-step text-to-image model tuned for the lowest latency and cost per image. On QuickSilver Pro it's $0.003 per generated image at market parity, billed from the same prepaid balance as your chat usage. Call it with the OpenAI SDK's images.generate against our base URL.
Try it
Runs on your own API key and balance, billed exactly like an API call.
1024 × 1024 for now — more sizes are coming.
Your image appears here.
The same request as code
curl https://api.quicksilverpro.io/v1/images/generations \
-H "Authorization: Bearer $QSP_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "sdxl-turbo",
"prompt": "A ripe red apple on a white marble table, soft studio light, product photo",
"size": "1024x1024",
"n": 1
}'At a glance
Cheapest, fastest text-to-image — previews and high-volume drafts.
Pricing (per generated image)
| Provider | Per image | vs QSP |
|---|---|---|
| QuickSilver Pro | $0.003 | — |
| typical hosted SDXL | $0.003 | same |
When to use
Reach for SDXL Turbo when you want the cheapest, fastest image path: preview thumbnails, quick iteration, and very high-volume passes where per-image cost dominates. Existing OpenAI-SDK image code works with a base-URL swap.
When to use something else
SDXL is an older architecture; for modern fidelity, prompt adherence, and in-image text, prefer FLUX.1 Schnell or FLUX.2 Pro. Advanced controls aren't exposed through the basic OpenAI Images parameters yet.
Quickstart (curl)
curl https://api.quicksilverpro.io/v1/images/generations \
-H "Authorization: Bearer $QSP_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "sdxl-turbo",
"prompt": "a sleek silver sports car, studio product shot",
"n": 1,
"size": "1024x1024"
}'OpenAI Images API. The response is OpenAI-shaped — data[0].b64_json holds the image. One key for chat and images.
FAQ
$0.003 per generated image on QuickSilver Pro, priced at market parity. Billed per image, from the same prepaid balance as your chat usage.
POST to /v1/images/generations with model="sdxl-turbo", or use the OpenAI SDK's images.generate. data[0].b64_json holds the base64 image.
Yes — it uses the OpenAI Images API. One key and base URL cover both chat and image generation.