Imagine a picture…

So I saw this neat project over at https://github.com/brycedrennan/imaginAIry which is a python program that makes stable diffusion more accessible and easy to use for common folk.

I’ve been playing with DALL-E 2 (more on that in another post), but this ImaginAIry tool just blew my mind.

You invoke it as

imagine “prompt”

and it invokes the stable diffusion to create an image that’s never been seen before that matches your prompt.

Seems hard to believe? Here’s some prompts that I gave it, and what it came back with

imagine “a crowd of people at a conference”

{“prompt”: {“software”: “imaginairy”, “prompts”: [[1, “a crowd of people at an AWS conference in black and white drinking coffee”]], “prompt_strength”: 7.5, “init_image”: “None”, “init_image_strength”: 0.6, “seed”: 346942399, “steps”: 15, “height”: 512, “width”: 512, “upscale”: false, “fix_faces”: false, “sampler_type”: “k_dpmpp_2m”}}

the example from the github page was:

imagine “a scenic landscape”

{“prompt”: {“software”: “imaginairy”, “prompts”: [[1, “a scenic landscape”]], “prompt_strength”: 7.5, “init_image”: “None”, “init_image_strength”: 0.6, “seed”: 534058548, “steps”: 15, “height”: 512, “width”: 512, “upscale”: false, “fix_faces”: false, “sampler_type”: “k_dpmpp_2m”}}

{“prompt”: {“software”: “imaginairy”, “prompts”: [[1, “a laughing child”]], “prompt_strength”: 7.5, “init_image”: “None”, “init_image_strength”: 0.6, “seed”: 321235062, “steps”: 15, “height”: 512, “width”: 512, “upscale”: false, “fix_faces”: false, “sampler_type”: “k_dpmpp_2m”}}

{“prompt”: {“software”: “imaginairy”, “prompts”: [[1, “a laughing child”]], “prompt_strength”: 7.5, “init_image”: “None”, “init_image_strength”: 0.6, “seed”: 93074687, “steps”: 15, “height”: 512, “width”: 512, “upscale”: false, “fix_faces”: false, “sampler_type”: “k_dpmpp_2m”}}

Just amazing… and it runs locally on my 2019 16″ MBP…