3.4 seconds in Fast mode
In the current 1024 × 1024 test on a MacBook Pro with M5 Max and 48 GB unified memory, Microsoft Lens Turbo took 4.0 seconds in standard mode and 3.4 seconds in Fast mode. Other Macs, sizes, settings, and LoRAs will change the time.
About a 16.8 GB download
The MLX 8-bit model needs at least 16 GB of unified memory and runs better with more memory available. Downloads can resume, and Radiant Canvas checks the model files before installation.
Private local generation
After the model is installed, generation can work offline. Prompts, results, seeds, settings, and Gallery records stay on the Mac. Microsoft Lens Turbo uses MIT and Apache 2.0 components.
Microsoft Lens Turbo is built for quick four-step generation. It is useful when you want to try several prompt ideas without a long wait between images. The model still runs on your Mac, so the prompt and generated image are not sent to a cloud generation service.
The MLX 8-bit download is about 16.8 GB and needs an Apple silicon Mac with at least 16 GB of unified memory. A Mac with 32 GB or more has more room for the model and other open apps. Model Manager shows the requirement before the download begins.
Speed is only one part of choosing a model. Use the same prompt and seed to compare Microsoft Lens Turbo with FLUX.2 Klein, Krea 2, Z-Image, or ERNIE-Image. Keep the result that fits the image rather than choosing from the benchmark alone.
Radiant Canvas
