Reference
Artinstring is provider-agnostic — every curated prompt is wired to the model that actually does the job best, not the one we have a deal with. Replicate runs are on us within your monthly quota; the rest are BYO key so the bill matches your usage.
Currently powering 113 curated prompts across the four providers below.
Hosted open-source models for image and video — Flux, SDXL, Kling, Veo and friends. The default backbone for our curated image prompts.
Fast inference for image, video, and audio models. Used for prompts that need lower latency than Replicate's queue.
GPT-4o, gpt-4o-mini and gpt-image-1 power our text prompts and a slice of the image lab.
Claude Sonnet 4.6, Opus 4.7 and Haiku 4.5 — the writers' end of the catalog (essays, code, briefs).
Every prompt page tells you which provider + model it dispatches to before you click Run. No surprises on your bill, no hidden fallbacks.