p-image

P-Image generates images using Pruna's optimized P-Image model family through the inference.sh CLI, invoked with the `belt` command. Pruna optimizes AI models for speed without sacrificing quality, and the skill exposes four models: P-Image (`pruna/p-image`) for fast text-to-image across multiple aspect ratios, P-Image-LoRA (`pruna/p-image-lora`) for custom styles via built-in presets, P-Image-Edit (`pruna/p-image-edit`) for image editing with multi-image support, and P-Image-Edit-LoRA (`pruna/p-image-edit-lora`) for stylized editing. It requires the belt CLI, installable via the belt CLI skill, and starts with `belt login`.

Core workflows are driven by `belt app run <app-id> --input '{...}'` with JSON parameters. Text-to-image takes a `prompt` and optional `aspect_ratio`; supported ratios are 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, or custom with width and height. LoRA styling adds a `lora_preset`, with presets including photos-realism, pixel-art, japanese-modern-look, cinematic-movie-style, graffiti-splash, neon-punk, anime-2-5d, ethereal-portrait, retro-90s-style, ink-sketchbook, and paper-cut. Image editing passes a `prompt` plus an `images` array of URLs; supplying multiple image URLs enables multi-image compositing such as collages.

Beyond the P-Image line, the skill points to other optimized Pruna models runnable the same way, including FLUX Dev, FLUX Klein 4B (noted as extremely fast at $0.0001/image), Qwen Image, Z-Image Turbo, and WAN Image Small for batch generation. Users can browse Pruna apps with `belt app list --namespace pruna` or `belt app store`. The documentation links to running-apps guides, an image-generation example, and streaming results for real-time progress, and it references companion skills for the full inference.sh platform, broader image generation, Pruna video, and FLUX models. The skill is scoped to the `Bash(belt *)` tool.