Draw Things is a cross‑platform app for iOS, iPadOS, and macOS that runs diffusion models locally, enabling offline AI image generation, infinite‑canvas outpainting, and image‑to‑video creation while keeping prompts and data private. It also provides on‑device LoRA fine‑tuning and an optional cloud compute tier for accelerated processing, with a subscription that unlocks higher cloud limits and custom LoRA uploads.
Funding
Funding not disclosed
Founders
Product
Problem
Creative professionals often rely on cloud‑based AI image generators that require an internet connection, expose proprietary prompts and source material, and incur latency or usage fees. Additionally, existing tools lack integrated features for animation, outpainting, and on‑device fine‑tuning, limiting iterative creative workflows.
Solution
Draw Things delivers a cross‑platform application that runs state‑of‑the‑art diffusion models locally on iOS, iPadOS, and macOS devices, eliminating the need for constant internet access and preserving user privacy. The app provides an infinite‑canvas outpainting engine, allowing creators to expand images seamlessly in any direction. Integrated image‑to‑video generation creates short animated clips directly from generated frames without external editing software. Users can train Low‑Rank Adaptation (LoRA) models entirely on‑device, keeping training data private while enabling personalized style control. For workloads that exceed local compute capacity, an optional cloud compute tier offers on‑demand GPU acceleration. A tiered subscription unlocks higher cloud limits, custom LoRA uploads, and additional privacy controls, while a free edition remains fully functional offline. All results are stored locally or transmitted over encrypted channels when cloud resources are used, ensuring end‑to‑end data security.
Target Audience
The primary users are independent designers, illustrators, game artists, and content creators who need private, high‑quality AI‑generated visuals on their personal devices.
Features
- Optimized on‑device diffusion inference leveraging Metal/Apple‑silicon kernels for real‑time image generation
- Offline operation with no data sent to external servers, guaranteeing prompt confidentiality
- Infinite canvas outpainting with seamless tiling and dynamic resolution scaling
- Image‑to‑video pipeline that synthesizes frame sequences using diffusion‑based video models
- On‑device LoRA fine‑tuning interface; model weights and training data never leave the user’s device
- Optional cloud compute layer for burst processing, accessed via secure API token
- Integrated model manager for downloading official diffusion checkpoints and uploading custom LoRAs
- Native UI across iPhone, iPad, and Mac with support for dark/light modes and touch/pen input