Introducing 'DALL E 2', a high-resolution, low-latency version of 'DALL E' that creates images from input text information
An improved version of the AI image generator 'DALL E' that automatically generates illustrations and photos from given words by combining natural language processing and image generation ' DALL E 2 ' has appeared. With high resolution and low latency, 'inpainting' has been added as a new function to specify and edit a part of an existing photo.
DALL · E 2
OpenAI's DALL-E AI image generator can now edit pictures, too --The Verge
https://www.theverge.com/2022/4/6/23012123/openai-clip-dalle-2-ai-text-to-image-generator-testing
'DALL ・ E' is a generator that generates illustrations and photos based on the phrases entered such as 'Pikachu riding a motorcycle' and 'Avocado chair'. Researcher Prafulla Dhariwal states, 'DALL-E simply applied the approach of the sentence generation language model GPT-3 to image generation.'
AI 'DALL ・ E' that automatically generates illustrations and photos from words such as 'Pikachu riding a motorcycle' and 'Avocado chair' --GIGAZINE
'DALL E 2' is based on the computer vision system ' CLIP ' announced in 2021 by the artificial intelligence research group
What are the characteristics of thinking of the image recognition AI 'CLIP' developed by OpenAI? --GIGAZINE
On the official website, a sample of what kind of image is actually generated by DALL E 2 is released.
For example, the words 'astronaut,' 'riding a horse,' and 'in a realistic style' produce the following pictures.
The composition seems to be running here.
It seems that it was made as such an art work, but it is automatically generated.
If this is 'Astronaut' and 'Riding a horse' in common and '
A sample of 'in painting' is also available. For example, the base photo is below.
Specify the area near the window, edit it, and add a sofa. You can do the same with ordinary photo processing, but the point is that it is automatically generated.
An example of processing the center of a room.
In addition, you can create variations from existing paintings. This is Vermeer's 'Girl with a
Variation example # 1. The composition and motif are the same, and the turban is a hat with a shape that I do not understand well, and it feels like 'I reproduced the picture I remembered vaguely'.
Variation example # 2. The turban remains blue, but left and right flipped. Earrings aren't pearls, they're something you don't understand.
Like 'DALL ・ E', this 'DALL ・ E 2' is provided for researchers and cannot be used by the general public, but OpenAI wants to be available in third-party apps. is showing.
Related Posts:
in Software, Posted by logc_nt