[ - ] symbolic 1 point 9 monthsJul 29, 2024 21:43:25 ago (+1/-0)
Runway Gen-3 is ok. Not local so you have 0 control over what it kicks out. Great model though, i wish it was open to the public to run through comfy. KLING and SORA both are around that level but also, not able to run locally. There are some Mid journey Loras and some of the XL models/PONY models that work well though for photorealism. The LCM tech is good as well, only about 5 to 10 steps for a cohesive image. Anything you experimenting with? Got any cool workflows you playing around with?
My workflow is I go to chat GPT and I say write me a story about xyz. Then I go to chat GPT and I say make me an illustration of the first paragraph of the story. Then I say make me an illustration of the second paragraph of the story. Then I say make me an illustration of the third paragraph of the story etc etc etc then you set those frames up into kling and you just get it to animate them.
Very cool, you should hang out at X0X7's forum, i post AI experiments sometimes and what i see happening on twitter in these regards. You ought to get into comfy UI, https://github.com/comfyanonymous/ComfyUI - its a node based approach to generating images and video. All local, so you dont have a company saying what political or satirical bullshit and memes you wanna produce are. Also, its fun! I've been playing with refeeding images back into it by using a sender and reciever node to automate that process, then i stitch it up in premier. Examples - https://files.catbox.moe/64d648.gif https://files.catbox.moe/czzay7.mp4
The image washes out unless you inject noise into it every now and then, which is something else i want to automate! ^_^ https://files.catbox.moe/k8w3nz.json you can just drag and drop the json script into comfy's workspace and you will have the workflow there for ya. I'm trying to do the same with but with video, currently it has the same issues of washing out over time. The big reason why i want to get that to work is then i can just hit go and take a nap for a few hours and wake up to a long 10+ minute video that is consistent across the chunks it generates. Also, it wont fry my GPU so badly because it can gen at these 32 or 64 frame chunks and then use FILM to interpet from chunk A to chunk B. If that makes sense.
[ + ] symbolic
[ - ] symbolic 1 point 9 monthsJul 29, 2024 21:43:25 ago (+1/-0)
[ + ] Crackinjokes
[ - ] Crackinjokes [op] 1 point 9 monthsJul 29, 2024 23:34:21 ago (+1/-0)
My workflow is I go to chat GPT and I say write me a story about xyz. Then I go to chat GPT and I say make me an illustration of the first paragraph of the story. Then I say make me an illustration of the second paragraph of the story. Then I say make me an illustration of the third paragraph of the story etc etc etc then you set those frames up into kling and you just get it to animate them.
Nothing I can show but that's the workflow
[ + ] symbolic
[ - ] symbolic 0 points 9 monthsAug 4, 2024 01:40:21 ago (+0/-0)
Examples -
https://files.catbox.moe/64d648.gif
https://files.catbox.moe/czzay7.mp4
The image washes out unless you inject noise into it every now and then, which is something else i want to automate! ^_^ https://files.catbox.moe/k8w3nz.json you can just drag and drop the json script into comfy's workspace and you will have the workflow there for ya. I'm trying to do the same with but with video, currently it has the same issues of washing out over time. The big reason why i want to get that to work is then i can just hit go and take a nap for a few hours and wake up to a long 10+ minute video that is consistent across the chunks it generates. Also, it wont fry my GPU so badly because it can gen at these 32 or 64 frame chunks and then use FILM to interpet from chunk A to chunk B. If that makes sense.
[ + ] Fascinus
[ - ] Fascinus 2 points 9 monthsJul 29, 2024 22:21:26 ago (+2/-0)
@Ror_Fly
Runway Gen-3 img-2-vid → Awesome.
My day is ruined.
#runwaygen3 #runwayml #midjourney
Direct link to embedded video: https://video.twimg.com/ext_tw_video/1818004235008516096/pu/vid/avc1/1080x1296/fYZxS4S3AXKE01Ow.mp4?tag=14
Mirrors of embedded video:
https://files.catbox.moe/gs0dx1.mp4
https://pomf2.lain.la/f/5c80p82u.mp4
https://cdn.videy.co/NYXqEK8D.mp4
https://qu.ax/tlLH.mp4