xororz/local-dream
Run Stable Diffusion on Android Devices with Snapdragon NPU acceleration. Also supports CPU/GPU inference.
What it solves
Local Dream provides a way to run Stable Diffusion image generation models locally on Android devices, reducing reliance on cloud services and allowing for offline generation.
How it works
The application leverages the Snapdragon NPU (Neural Processing Unit) for hardware acceleration via the Qualcomm QNN SDK, though it also supports CPU and GPU inference. It currently focuses on SD1.5 and SDXL models, utilizing libraries like MNN for CPU execution and tokenizers-cpp for text processing.
Who it’s for
Android users with Snapdragon-powered devices, particularly those with Hexagon V68 architecture or newer for SD1.5, and Snapdragon 8 Gen 3 or newer for SDXL.
Highlights
- NPU Acceleration: Specifically optimized for Snapdragon NPUs to improve performance.
- Model Support: Compatible with SD1.5 and SDXL models.
- Upscaling: Integrates Real-ESRGAN and UltraSharpV2 for image upscaling.
- Local Execution: Runs entirely on-device, supporting CPU, GPU, and NPU inference paths.
Related
- Project
- Project
- Project
- Project
- Project