这样也许不涉及版权了？而且网站还提供了有限的编辑功能。现在感觉还不是很完善。图片上只要涉及文字基本是有点乱。
I tried playing DALLE.2 to generate pictures for some topics on selling computer parts. Some of the topics can produce reasonable photos, but some of them make less sense. Do not try to add TEXTs to the pictures because I did not see any success.
I. This picture was generated for a topic of selling used GPU. The input texts are “game GPU graphics card installed in a game machine at home blue lights.” Although it does not match what I expected, the picture is still usable. It won’t create much confusion.
II. I want to generate a picture with many server CPUs to match the topic of “selling used computer processors”. The generating texts are “Intel High-End Server CPU in quantity”. The shape of CPUs looks like an Intel processor and contains an array of CPUs. But these processors do not look like high-end ones. I can still use it.
III. Now let us look at the generated results of texts “ddr4 pc4 server memory”. I expect it can produce some details on the memory. At least I can tell if they are DDR4 or not. Among all the returns, this one looks much better. But not what I expected. I tried different text descriptions, but no good results returned. I may use it for the “selling used RAM” topic.
V. The worst results are related to the topic of used SSD hard drives. None generated images are meaningful. I do not know where the system acquires the training data. I think, as long as the system has crawled enough training examples, it should generate good examples. It is only my guess, of course.