What are the best online tools for AI-enabled image creation or editing that have more interactivity? So, not just "you write text, wait 30 seconds, image comes out", but more two-way feedback between user and diffusion algo, ideally without 30 second latency at each step.
civitai, leonardo come to mind. Allowing more control with things like controlnet, img2img
Steve McDonald is worth looking up, Toronto based artist who augments his “analog” capabilities with different ai tools. Unique style, using tools for augmentation more than generation, interesting to see. Think he has tutorials/course https://instagram.com/stevedmcdonald?igshid=MnIxdm0xcHByNmM2
If text-based interactions, DALL-E 3. If you UI-based, https://playgroundai.com/
If text-based interactions, DALL-E 3. If UI-based, https://playgroundai.com/
Maybe it's too simple, but using gpt-4 with Dall-E 3 does the trick. The two-way feedback is nice, allowing me to iterate using text. Still, when I make changes, the modified image loses important parts of the original.
Love this question. Feel like this will play such an essential role in the creation of short form content. What’s the intention you have in mind for how you would want to use the output?
Unfortunately online tools are quite limited in that regard. Your first, and best option is installing AUTOMATIC1111 and then customize it with plugins: https://github.com/AUTOMATIC1111/stable-diffusion-webui (but it takes a lot of time) Thread..
create something and bring it to life! I've been experiencing with Luma lately and it's really powerful if you're looking into 3D models as well https://lumalabs.ai/
LCM - I really like this example from M. DesLauriers (offline and almost real-time) https://twitter.com/mattdesl/status/1719501171814273258
Wow, that's amazing! 😃 Thanks for sharing this with us. It's inspiring to see your [achievement/experience/story]. Keep up the great work! 👏👏
Photoshop. For real. In painting is amazing and useful and the best application of imagegen right now.
Start a stable diffusion colab instance Set it up to output small images, at frequent step sizes, running on a powerful gpu (~ $10 / mo) Here's a general stable diffusion notebook but better exists if you hunt https://colab.research.google.com/github/huggingface/notebooks/blob/main/diffusers/stable_diffusion.ipynb
Weak answer as this doesn't solve the interactive part, but some form of LCM Stable Diffusion setup on Google Colab or HuggingFace is the best online setup you can get right now. When imagegen takes only a few hundred milliseconds, iterating is fast.