Skip to content
note · kiryu_mizuki · 图像视频创作分享·· 2026-10-16AI score22

Running AI image generation on a 6GB VRAM laptop: INT8 practice with Krea2 Turbo and Qwen-image2.1

おいらのやり方はコレだよ♪

AI briefing

Creator Shuihui runs AI image generation locally on an RTX 3060 Laptop (6GB VRAM), generally choosing INT8 versions of models rather than GGUF versions, considering the latter's image quality not to their taste. Currently, generation with Krea2 Turbo plus 4STEPLORA takes about 70 seconds, Qwen-image2.1 with the official int8 model and 8STEPLORA takes about 65 seconds, and two reference images take about 70–80 seconds.

Open original note.com