Draw Things
Draw Things is a local AI image generation application optimized for Apple Silicon. It supports local inference with Stable Diffusion series models and includes a built-in JavaScript sandbox environment that supports automation scripts for batch image generation.
Timeline (last 90 days)
The Draw Things software already supports the Qwen Image 2.1 model.
Searching for Qwen Image 2.1 in Draw Things shows the original teacher model and the Turbo accelerated model, each offered in five quantization precisions; the version without a suffix is the 8-bit version, and the one with Exact is the FP16 version.
The first-and-last-frame version of MiniMax H3 (FLR VA) supports three modes—text-to-video, image-to-video, and first-and-last-frame video—making it the first model on Draw Things to support first-and-last-frame generation.
The Talmate 3-step acceleration LoRA version optimized by KeyJay is only 182MB in size and can be imported into Draw Things to run.
MiniMax H3 is now officially supported on Draw Things, allowing Mac and iPad users to run the model on local devices.
The UP host conducted a local real-world test of MiniMax H3 on Draw Things using an M5 iPad Pro with 16GB of memory.
When Draw Things supports audio plus images as joint input conditions, it is expected to enable the creation of short films exceeding 15 seconds with coherent audio consistency.
It is advised to first run through the workflow at a small size (e.g., 576×384) with fewer frames (e.g., 107 frames), and then gradually increase the size and number of frames.
On Draw Things, MiniMax H3 is available in two types and eight quantization precisions.
All Facts (12)
Mac mini上本地生图单张耗时30秒到2分钟
75%UnverifiedDraw Things has a built-in JS sandbox environment that allows users to write automation scripts to invoke the image generation pipeline for batch generation.
60%UnverifiedDraw Things is a local AI image generation application optimized specifically for Apple Silicon (M-series chips), supporting local inference for the Stable Diffusion series of models.
60%UnverifiedSearching for Qwen Image 2.1 in Draw Things shows the original teacher model and the Turbo accelerated model, each offered in five quantization precisions; the version without a suffix is the 8-bit version, and the one with Exact is the FP16 version.
50%UnverifiedThe Talmate 3-step acceleration LoRA version optimized by KeyJay is only 182MB in size and can be imported into Draw Things to run.
50%UnverifiedOn Draw Things, MiniMax H3 is available in two types and eight quantization precisions.
50%UnverifiedIt is advised to first run through the workflow at a small size (e.g., 576×384) with fewer frames (e.g., 107 frames), and then gradually increase the size and number of frames.
50%UnverifiedThe first-and-last-frame version of MiniMax H3 (FLR VA) supports three modes—text-to-video, image-to-video, and first-and-last-frame video—making it the first model on Draw Things to support first-and-last-frame generation.
50%UnverifiedThe Draw Things software already supports the Qwen Image 2.1 model.
50%UnverifiedMiniMax H3 is now officially supported on Draw Things, allowing Mac and iPad users to run the model on local devices.
50%UnverifiedThe UP host conducted a local real-world test of MiniMax H3 on Draw Things using an M5 iPad Pro with 16GB of memory.
50%UnverifiedWhen Draw Things supports audio plus images as joint input conditions, it is expected to enable the creation of short films exceeding 15 seconds with coherent audio consistency.
50%