Player not loading? Watch on YouTube
This tutorial walks through image generation, inpainting and image-to-video creation in Draw Things on a Mac. The speaker describes the image workflow as offline and private after downloading the models. Downloads can occupy several gigabytes, and the first generation may take longer while the app loads a model into memory.
For text-to-image generation, the speaker selects Z Image Turbo, a square resolution and the recommended settings, then enables Real-ESRGAN x2 upscaling. The inpainting example replaces an iPhone advertisement on a billboard with a message about responsible phone use. The speaker masks the billboard with the eraser tool and uses Ernie Image Turbo; the surrounding street scene remains intact in the demonstrated result.
The video example pairs high-noise and low-noise 2.2 I2V models, with the latter acting as a refiner. It animates a train-station image at 448 × 768, using 16 frames per second and 33 frames. The speaker cautions that video generation demands more resources than a single image and that speed depends on the Mac and model.
The pricing discussion describes a free edition for local AI image generation and a Draw Things Plus plan with cloud compute. Larger models, higher resolutions and less powerful hardware can lengthen generation times.
