HexAnima fork runs offline image generation on Snapdragon 8 Gen 2 NPU at ~7s/image
ilinkys · reddit · 2026-11-17
A developer forked HexAnima to bring fully offline AI image generation to Snapdragon 8 Gen 2 (Hexagon V73) devices, no root required.
- Generation: 2B/2.9B Anima models, on-device safetensors conversion, live previews, adjustable steps/CFG/samplers/seeds, negative prompts and prompt weighting
- LoRA support: Kohya, Diffusers/ComfyUI, LyCORIS LoKr/LoHa formats with adjustable weights and q8 caching
- Civitai integration for in-app downloads; img2img, inpainting, HiRes Fix, ADetailer face refinement, queues, comparison grids, presets
- Temperature-aware cooling mode throttles the NPU when hot
- Benchmarked on Poco F6 Pro (12GB, 512×768): 2B Turbo 4 steps 7s, 2.9B Turbo 10–12s, 2B non-Turbo 20 steps 85s
- Requires Android 10+, 1.2GB shared files, 3.6–4.2GB per model; open-source APK released
More from Infra
- Together AI and Ollama explain how prefix caching keeps long agent sessions fast — togethercompute · 2026-11-17
- Microsoft Outlook, airports, major US banks hit by outages in quick succession — robleclerc · 2026-11-17
- Nvidia's Ian Buck marks two decades of CUDA, from one 2006 toolkit to a library for every workload — firstadopter · 2026-11-17
- NVIDIA's cuPotential speeds molecular dynamics over 3x on Blackwell, eclipsing a decade of gains — AllThingsApx · 2026-11-17
- ETH's Hoefler at SC26: Networks Matter More Than GPUs for AI Supercomputers — thoefler · 2026-11-17
- Resend latency drops across the board: p50, p90 and p99 all trending down — philnash · 2026-11-17