#3737th of 89 in Image and videoFor image, video and design, image and design
local-dream
Android app that runs Stable Diffusion on-device, with Snapdragon NPU acceleration
- Stars
- 4k
- License
- custom license
- Last commit
- Nov 2026
- Last release
- Jul 2026
- Language
- Kotlin
2,227 on 25 Apr 20264,048 on 14 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Local Dream is a Kotlin Android app that generates images with Stable Diffusion 1.5 and SDXL entirely on the phone. It uses the Qualcomm QNN SDK for Snapdragon NPU inference and MNN for CPU/GPU fallback, and includes Real-ESRGAN and UltraSharpV2 upscalers plus an NSFW filter. Guides and documentation live on a separate site.
Who it is for: Android users who want offline image generation on Snapdragon phones
Strengths
- NPU inference on Snapdragon: SD1.5 on Hexagon V68 or newer, SDXL on 8 Gen 3+
- CPU/GPU fallback through MNN for devices without a supported NPU
- Built-in upscaling with Real-ESRGAN and UltraSharpV2
- Runs fully on device, no server required
Weaknesses
- NPU path limited to Snapdragon; other SoCs only get CPU/GPU
- SDXL needs substantially more memory than SD1.5
- Only SD1.5 and SDXL are supported; SD2.1 is no longer maintained
- License is not a recognized standard identifier (NOASSERTION); check before reuse
What it needs
- GPU optional
- Needs Qualcomm QNN SDK, MNN
- Models: SD1.5, SDXL
What changed
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Also in Image and video
See all 89| Rank | Project | Score |
|---|---|---|
| 1 | frigateLocal NVR with realtime object detection for IP cameras | 93 out of 100 |
| 2 | MoneyPrinterTurboGenerates short videos from a topic, with script, footage, voiceover and subtitles | 88 out of 100 |
| 3 | toonflow-appInfinite-canvas tool for AI short drama and comic video production | 88 out of 100 |
| #4 | InvokeAIWeb UI for diffusion image generation with inpainting canvas and node workflows | 87 out of 100 |
| #5 | ComfyUINode-graph editor for running diffusion image, video, audio and 3D workflows | 85 out of 100 |