SDAI 1.1.0

SDAI 1.1.0 is a feature release focused on local generation, provider choice, device insight, and practical usage tools across Android and iPhone.

Highlights

• Added an on-device benchmark for local generation guidance.
• Added Android Local SDXL generation through stable-diffusion.cpp.
• Added iPhone local generation through Silicon Diffusion Core ML.
• Added iPhone local generation through Silicon Diffusion PrismML Bonsai.
• Added Fal.ai and ArliAI provider support.
• Added storage usage and network usage screens.
• Improved provider selection with search, filters, sorting, readiness labels, and feature tags.
• Improved iOS gallery loading and background-work handling.

Local Generation

This release expands SDAI's local generation story on both platforms:

• Android Local SDXL uses stable-diffusion.cpp for on-device SDXL-compatible text-to-image generation.
• Silicon Diffusion Core ML brings Core ML Stable Diffusion generation to supported iPhones.
• Silicon Diffusion PrismML Bonsai adds a beta PrismML Bonsai Image 4B / MLX local generation path for supported iPhones.

Local generation remains device- and model-dependent. The new benchmark is designed to help users understand whether their device is a good fit before committing to heavier runs.

Providers

• Added Fal.ai hosted generation support for compatible text-to-image and image-to-image endpoints.
• Added ArliAI SDNext-compatible hosted generation support, including checkpoint discovery, sampler selection, steps, CFG scale, negative prompt, and batch generation where supported.
• Improved the provider setup experience with search, type/readiness/tag filters, sorting, and clearer provider labels.

Usage Tools

• Added storage usage screens for cache, gallery, and downloaded model files.
• Added network usage screens for model downloads, configuration downloads, and provider inference traffic.
• Added local benchmark result storage and recommendation flow for supported local providers.

iOS

• Added Core ML and Bonsai local-provider runtime integration.
• Improved gallery loading performance.
• Added iOS background-work support for long-running generation flows.

Android

• Added stable-diffusion.cpp based Local SDXL provider integration.
• Added model-catalog and local-model selection support for SDXL workflows.
• Added benchmark and usage-tool platform actions.

Notes

• Local AI generation can be memory-heavy. If a model is too large for the device, try a smaller or quantized model and check the benchmark recommendation first.
• iOS local providers are intended for supported iPhone devices. This release does not announce iPad support.
• SDAI remains open source, ad-free, telemetry-free, and provider-first.

Download and project links:

• Website: https://sdai.moroz.cc
• Full Changelog: 1.0.0...1.1.0