note · akb428 · 图像视频创作分享·· 22 days agoAI score51
Internet café real-world test: RTX 5060 running ComfyUI image generation — how fast are Anima and Qwen-Image 2.1?
[ネットカフェで生成AI] RTX 5060でComfyUI画像生成はどこまでできる?ネットカフェマンボーのeスポーツルームで試してみた(Anima・Qwen-Image 2.1)+ローカルLLM小説【2026秋】
AI briefing
The author tested RTX 5060 (8GB VRAM) running ComfyUI image generation: Anima took 57 seconds per image, while Qwen-Image 2.1 reference generation took 47 seconds per image. The local LLM (Ollama's gemma4:12b-it-qat) managed only 14 tok/s, becoming the speed bottleneck when used together. The author also tested their own novel-generation tool Inkubus: about 14 minutes for 10,000 characters, but the 12b model's quality is limited.
Open original note.com
Image GenerationComfyUI & WorkflowsTutorials & Practice#Tutorial#Image generation#ComfyUI & workflows#Alibaba