Next, add Segment Anything…
2022-04-11 Really impressive animation
https://www.reddit.com/r/StableDiffusion/comments/12ii0ob/hi_every_budy_i_remake_ive_new_song_to_animation/ https://www.youtube.com/watch?v=oUe8ZyD7-ww
2022-04-11 Face
https://www.reddit.com/r/StableDiffusion/comments/12ht3g1/animated_stuff/
2022-04-11 Depth Extension
https://www.reddit.com/r/StableDiffusion/comments/12ijx7o/zoom_and_slight_rotation_using_keyframes_drawn/ https://www.reddit.com/r/StableDiffusion/comments/12ik1j7/i_posted_a_few_day_ago_about_how_to_animate_using/
2022-04-11 Animation
https://www.reddit.com/r/StableDiffusion/comments/12i9qr7/i_transform_real_person_dancing_to_animation/ sd-webui-mov2mov https://github.com/Scholar01/sd-webui-mov2mov https://www.youtube.com/watch?v=n5-PXi9UGJk
2022-04-05 Cohesive Animation in Stable Diffusion Img2Img
https://www.reddit.com/r/sdforall/comments/12bpbqp/cohesive_animation_in_stable_diffusion_img2img/
2022-03-31 Openpose + TemporalNet
https://twitter.com/toyxyz3/status/1641599874448891905
2022-03-28 - AI Animation w/ SD workflow
https://www.reddit.com/r/StableDiffusion/comments/1240uh5/video_to_anime_tutorial_full_workflow_included/ https://www.youtube.com/watch?v=kmT-z2lqEPQ
2022-03-26 - SOTA research summary
https://twitter.com/ptsi/status/1639828628547465217
2022-03-23 Tips for Temporal Stability, while changing the video content
2023-03-21 Workflow
https://aidreams.tech/blog/f/deforum-basics-with-stable-diffusion https://www.reddit.com/r/StableDiffusion/comments/12hqlxi/animated_music_video_deforum_ffmpeg_and_clipchamp/
Basic workflow:
-render out every 2 frames of a 24fps video (we convert the 12fps back to 24fps later)
-img2img with low denoise 0.3~0.4 works for me
-choose a high resolution for the image in my case it was 1024x1600 ish
-I choose HED and leave the values at default
-test out different frames to see if things look consistent
-save the seed
-run batch mode
-Put the 12fps to convert to 24fps using Flow Frame
-Put the flow frame footage into Davinci resolve and use de-flicker in florolight mode
*No fancy tricks on this one but the most time consuming part was finding the right resolution vs denoise value to give me something that looks different enough but also consistent from frame to frame plus enough detail in the face.
*Then it took 24 hours to render out the de-flicker from resolve. Definitely insane amount of RAW computing power but it did the job.
https://share.hsforms.com/16C8k9E5FR5mRLYYkwohdiQ4sk30
Leonard Lin
I’d like to be able to get a daily summary of my news (Twitter List, Reddit, HN) feeds. I’d also like to be able to pull in real-time financial data and do some analysis on that. And to be able to send tasks/note directly to my Notion, Clickup, Calendar, etc.
I have started to build some toy experiments in Langchain this past week against both the OpenAI API and local LLMs. See: Hi what version of GPT are you based off of?