🎥｜animation

[oak urchin]

yeah, sure, same here. But i'm just talking about one axe at the time

I can't visualize yours, just on one axe

I can visualize mines, i'm pretty sure one will feet, that's my point

*more the values inside, than the equation themself, but still

[vast mural]

i understand why i have formulas in y-rotation and values in x and z translation

but i dont get the right values 90 degrees into the frame

i will try more this we

[oak urchin]

Cool, let's keep the contact (:
the first one show the progress

[oak urchin]

Last ping, I spent some time to produce some different approach, to run and see. As basis.
This one is the first i'll try, the more logical that i can visualize :
"angle": "0:(0)", "zoom": "0:(1)", "translation_x": "0:(cos x (t x 2 x pi/900))", "translation_y": "0: (0)", "translation_z": "0:(sin x (t x 2 x pi/900))", "transform_center_x": "0:(0.5)", "transform_center_y": "0:(0.5)", "rotation_3d_x": "0:(0)", "rotation_3d_y": "0:(t x 2 x pi/900)", "rotation_3d_z": "0:(0)",

Not fixed values here, didn't test
Also replaced * with x, sry lol

a very slight angle can be visually awesome, as i see it

[vast mural]

rotation_3d_y should be a constant, as this is defined in angles per frame. For a constant rotation, it should be same all 360 lap

[oak urchin]

Just turning slighty the camera, like a plane going left or right

[oak urchin]

Yep, probably, just a draft. The equation/syntax is the point (:

[vast mural]

900 frames?

[oak urchin]

just take the idea, lmao

[vast mural]

transform_center and zoom are not used in 3d I hope

[oak urchin]

This is a circular motion, around a point in the middle, for me

that's also a phrasing for my point, whatever the values

The distance is from the prompt/picture, seems a good approach to me

[oak urchin]

Sure, the rotation around the axe will take 900 frames to be complete, 1 full rotation. You can't imagine it !? I do

We need a math equation/function visualizer, we would see the motion

[vast mural]

actually just getting a printout of the resulting deforum values per frame would help.

just discovered that translation z is not really relevant for the rotation...

here the cos formula is way to frequent and small, but the cat rotates right anyway... weird

[oak urchin]

lol, indeed.

And how far i don't understand your equation

[twilit shale]

https://www.youtube.com/shorts/i-5qs6kQO7w

[polar pivot]

hey! how you make it that 3d? with ehat parameters? possible in parseq?

[vast mural]

Just straight deforum with 3D rotation

https://github.com/RobertSvebeck/StableDiffusionThings/blob/main/deforum_settings/object_rotation/20230607202141_settings.txt

[vast mural]

just another comment, this particular video is stiched from two videos because currently I can only render 90 degree rotation.

[hot dew]

How are you guys able to produce such clean results? I am still struggling with removing flicker

[vast mural]

Having used Deforum for just a few weeks, I am absolutely no expert. But my first go-to things for removing flickering is trying higher values in Cadence and higher strength schedules.

[hot dew]

Thanks. I try cadence 5, but it sometimes generates an error on Automatic1111. So I usually set it to 3.

How much do you use?

[vast mural]

The turning cat up there has cadence 2 and strength 0.8

your video there has a video as input?

[hot dew]

yes

[vast mural]

havent tried that yet 🙂

[hot dew]

Video Input doesn't have cadence option

[vast mural]

will try with the cat video as input and see what I find.

[hot dew]

But your video is amazing without any input. You didn't use image input either?

[vast mural]

Thanks. No, this is plain prompting. No inital image.

[hot dew]

Amazing

[vast mural]

Seed = Iter is also good for lower flickering. Also good to find a good starting seed and then stick to that for each try. Makes it easier to understand whats changing.

[hot dew]

Thanks. Will keep that in mind

[vast mural]

Video input is a mess! 🙃 impossible to get any good results. Perhaps with controlnet?

[hot dew]

I am using Camenduru's notebook. Just Controlnet doesn't work in that in Deforum

[cobalt thunder]

https://www.youtube.com/watch?v=w-VXY8nbjF0

[faint rivet]

https://www.youtube.com/watch?v=eOuSdUsDkvs

[cloud smelt]

https://youtube.com/shorts/X1ensXdtWJQ?feature=share

[dense crater]

https://youtu.be/Lc01cAR2Fj0

[brittle rivet]

https://youtube.com/shorts/99gicsCL5eI?feature=share

[vast mural]

rotating a dragon was not as easy as a kitten….

[vast mural]

@oak urchin I think 360 rotation is not possible. Look at this animation, where I start from a sideview and rotates +90 towards to the front. It just collapses, so I am guessing that the model does not understand what to do. Will try sd2.0 and see if it works better.

[vast mural]

360 rotation not working, but at least I can now change rotation direction half way and "waggle" from side to side. 🙂 This is just one prompt, one render, no post production.

[oak urchin]

The model can't draw the rear, if your prompt ask for the front, for sure.

[vast mural]

YEs but the model can draw the front if it starts with a profile picture either, even if the profile is not described in the prompt. It seems some seeds are just impossible to rotate also. Sometimes a seed refuses any rotation

Look at this. Same deforum settings, just this seed refuses to rotate

wrong video

I menat this

[copper vector]

I saw this amazing reel, I wanted to know if Stable Diffusion was used or could be used to make something like this? Thank you so much !

[vast mural]

you mean the anime clip in the beginning?

[copper vector]

There are two clips, 'his childhood photo to anime clip' and 'the sinking in water to house animation'

[copper vector]

I wanted to enquire for both

[vast mural]

I think the first one is possible but the second one probably not

[oak urchin]

Gimme real Text to video peeps, process a video seems soooooooooooooooooooooooooooooooooooooooo boring to me. And easy as f***

[nova hazel]

Hi, which GPU are you using? how much RAM

[vast mural]

Nvidia A2000, 12gb VRAM

[zealous linden]

could the techniques of tecogan be used to create more consistent animations?

[safe oriole]

https://youtu.be/0Vzk7HxrYck

[plucky plover]

Lovejoy - Portrait Of A Blank Slate (AI Animation) https://youtu.be/xrGreJE7FsA

[fallen girder]

Neuromancer (Text2video / RunwayML Gen-2) [4K]
https://youtu.be/ta7_oGaZNXI

[ebon sparrow]

Control net M2M TEST - 1

[cobalt thunder]

https://www.youtube.com/watch?v=LuCXo0uuBOE

[static niche]

is there a tool/addon/module for creating a transition animation from one image to another?

for instance, changing a dog into a car or a baby into an adult

[faint rivet]

https://www.youtube.com/watch?v=XxUSMeBzZow

[spare coral]

Hey guys! We (Meat Dept.) just published this music video made with Blender, Octane and Stable Diffusion. Can’t wait to have your feedbacks.. Enjoy!

https://vimeo.com/835172307

[cloud smelt]

https://youtube.com/shorts/_O2st15jbR4?feature=share

[vast mural]

The guys who work with the Auto1111 InfiniteZoom extension is working on such a thing. There is a branch with start image and exit image.

[fallen girder]

AI Generated Song (SongR + D-ID)
https://youtu.be/X8tm-I6w9k4

[hot dew]

this looks like runway

[crystal gust]

Freeze last frame, use controlnet and generate multiple images with different prompt. Then compile them in post in video editing software.

[thick sable]

Basic video editing color curves, used some HUD stock footage and blended it over my footage and threw it into warpfusion !

[faint rivet]

https://www.youtube.com/shorts/KV3nOYsql1E

[static niche]

Thanks, will dive into it. Appreciate the response.

[ruby nova]

God damn gimme promy 😃

[ebon sparrow]

Thank you♥

[somber geode]

So controlnet can do the frames with diff prompt in batch or should i take each frame and process it through img2img then put them together?

[crystal gust]

controlnet has batch process option but thats not how its made

A single frame on which the original video stops is taken, then put in CN and generate img with prompt, change prompt generate new image and so on

[crystal gust]

I think there is an option called dynamic prompting in which you can put multiple prompts at once and it will generate images based on each of those prompts

[somber geode]

Ok thank u

[somber geode]

do u think this type of video is good?

[crystal gust]

Good as in the original video to make such video or the final result video?

[young forum]

Did a small test https://vm.tiktok.com/ZGJQgDPPp/

[dull shard]

Hey all, I've just gotten into warpfusion as I'm an editor who's been tasked with a music video. I really love this reference and want to incorporate the look and feel into band performance footage. This video has ever-changing visuals that seem to follow some simple lines of motion. Would this be achieve by a single run, or is it multiple runs of Warpfusion that have been seamlessly woven together in an edit? Any help would be appreciated - https://www.instagram.com/p/Cnl9vOcIe6h/

never_ever_never_land

The Light Painter

Likes

164

Instagram

[zinc scarab]

Hi guys. Im getting an error when using Deforum. Just installed today so no im not really familiar with it. Is this the right room to ask?

[vast mural]

There is a dedicated Deforum server

[vast mural]

https://github.com/deforum-art

[strange sierra]

Video-to-video outcome

[cloud smelt]

https://youtube.com/shorts/n53zX4i96aU?feature=share7

[solar hearth]

https://www.youtube.com/shorts/K6q2uOlcHE8

[dusk island]

Ayooo, clutch. Thank you for that.

[low nebula]

Heya you have any advice for getting frame to frame consistency? Reddit is in shutdown. My results are more erratic.

[faint rivet]

“Aᴍᴇʀɪᴄᴀɴ MAGA” 🐸 Rᴇᴛᴜʀɴ ᴛᴏ Sʜᴀᴅɪʟᴀʏ 🐸 ᵀᴿᵁᴹᴾᵂᴬⱽᴱ ²
https://www.youtube.com/watch?v=9Q2GmOR64tg

[twin shadow]

https://cdn.discordapp.com/attachments/1025266140445933648/1119387253370794075/hey__.mp4

[lilac pier]

woah, may I ask how this was done? it's very fluid and I thought it was a real anime for a moment!

[twin shadow]

Thank you 🙏 I used img to img batch rendering

[untold pulsar]

https://youtu.be/H7fn1ej3bkk
"Nightmare Chronicles & Stable Diffusion: A Match Made in Horror Heaven"

[lilac lion]

Hey guys, If I have an image, how could i generate an video for this such that the painting style of the video is the same as the that of the image? Thank you guys!

not from a txt to video, I hope the input image can become one of the key frame of the generated video

[quaint night]

Hey guys could anyone share me the github example if anyone generated the animation using stability AI

[ebon sparrow]

Control net M2M TEST - 2
https://youtu.be/AO6Cr6_ze7A

[toxic glen]

Awesome! I have found a new home!!!

[toxic glen]

I also love to make horror type animations!

[solar hearth]

https://youtube.com/shorts/avdtxeSxFPk

[toxic glen]

Duuuude sick!!!

Check this bad boy out

[solar hearth]

I gotta try more guided images, it's awesome.

[toxic glen]

I have literally no idea how you created the one you posted tho, I can't say I've ever seen anything like it

Oh, I've never used guided images before.

[solar hearth]

The transitions were so smooth, it fooled me 😄

[solar hearth]

The source video is from environmental tracking in 3D software where an exploding particle bubble is created in the field, and then Deforum/ControlNet on top of it to create the exploding nebula, with some post in Davinci Resolve.

[toxic glen]

Super dope man. I'd love to pick up some pointers from you.

[faint rivet]

https://www.youtube.com/watch?v=qYAI6iJhV8c

[faint rivet]

Trying out a new way of adding optical flow in deforum - after generation. It looks promising!
This is strength 0, using two controlnets,

• CN lineart/lineart on loopback with the Prompt is more important. Weight 0.5
• CN no preprocessor/tile set to the video file with Controlnet is more important. Weight 1

[safe oriole]

thanks for the workflow 🙂 and of course your deforum automatic work

[faint rivet]

I added hybrid video to the notebook first! But yeah, I did a lot more with webui.

For reference, here's the original

[safe oriole]

waiting to see if we get some coherence on 'step' functionality, eg, diffuse then coherence , or diffuse non mask coherence inside mask , probably end up being a llora built inside of a render which might be a bit insane 😄

[fringe carbon]

@ebon sparrowyo gimme that code and i'll give soething equally aluable in return. or i can pay $.

yoooooooo

i selling a service

ro improve your black metal logo reasonalby fast

[toxic glen]

https://civitai.com/articles/606/dwh-2-animating-your-own-images-with-deforum

[wraith flame]

Made with Stable Diffusion and Ebsynth.

[ebon sparrow]

Thank you for your interest in me.
My videos are not generated by code, but by a simple method that uses img2img output based on real images.
If you still want, we can show you how to make it or create a video set to music!

[fallen girder]

"Summer Dreams" An AI generated short film made with Deforum
https://youtu.be/H67UtjzlqmM

[solar hearth]

Made with free software, SD/Deforum/ControlNet!

[ruby nova]

plz rotation Amigo

[toxic glen]

I'm sorry?

[quaint night]

Which one did you use warp fusion, disco diffusion, or stable diffusion Animation API

[wraith flame]

Stable Diffusion, Deforum and ControlNet Tile. Model: Inkpunk Diffusion.

[quaint night]

Have you tried stablity APIs, do we get this kind results

[wraith flame]

The first sequence is made with ControlNet Canny.

[wraith flame]

sorry I don’t know what is that

[quaint night]

I need some guidance to generate like you

[wraith flame]

I made it with a tutorial but I don’t know if I can share it here

[warped portal]

Hey fellas, first time here so i dont know really where to post it..
some one know how to make those image to animation thing?
But not the one that changing frames, the one that you take an image and make it into a 3d depht thing...

Example- https://www.instagram.com/p/CtlRrA7LJDO/

digitalferno

Disney - Mickey Mouse

#mickeymouse #cartoon #animation #digitalportrait #cartoonrealism

Instagram

[quaint night]

Check your DM

[mossy ridge]

https://m.youtube.com/watch?v=DAhw6WjNoTM&t=3049s

[fringe carbon]

That'd be great! Thank you so much!

[ebon sparrow]

https://youtu.be/fNvo4f6CWro

[livid wasp]

https://youtube.com/shorts/yp1wv_n_iiU

[pliant swallow]

Created with automatic 1111 https://youtu.be/FXl_1NHJIcA

[tough edge]

hhow are yall animating stuff lmao

[toxic glen]

https://civitai.com/articles/554/dwh-1-creating-your-first-animation-with-deforum-in-1-click

[tough edge]

ty

[hot dew]

Hi,
Has anyone tried animating AI gen images to tell a story? Perhaps used them in OpenToonz or anyother software... Is that possible?

[fringe carbon]

I might just be a step closer.

https://pastebin.com/cZxiFJXr -- feedback and constructive criticism welcome!

Hoping there's a way to do some trickery in between the frames. Maybe combine it with various other tools.

Anybody worked on anything similar?

[gilded bane]

Used Kaiber to make a music video for a goofy lil song I made

[hot dew]

wow! which tool did you use?

[nimble hearth]

https://github.com/kabachuha/sd-webui-text2video/discussions/132

[hot dew]

thanks... checking out right away

[hot dew]

This is amazing. Going through the workflow, I couldn't help but think, that if something similar can be possible in deforum also? It's probably a dumb curiosity of my non-technical mind, But just putting it out here to curb it.

[nimble hearth]

https://github.com/damo-vilab/videocomposer

[hot dew]

Thanks... Will check out. You are amazing

[fringe carbon]

hmmm

Maybe it's easier to just put it in a simple Ruby script

[fringe carbon]

Ruby's like Python but non-technical minds like us can understand. I'm trying to combine all of this into a neat little Ruby script. Anyone wanna join and get to enjoy the finished product?

We'll be using a lot of extra cinematography techniques as well.

[hot dew]

Would love to try the finish product. but the linnk you shared went over my head 😂

[fringe carbon]

One sec. it uses two parts, 1. setup a dreambooth model and generate a bunch of random artist promo photos of exceptional quality. and then this 2. comes, reuses any dreambooth model together with ruby's amazing libvips. libvips will also do professional color grading, making instagram filters look like a joke.

[inner crest]

Time to show the community what ai art is...

https://tenor.com/view/stretch-stretching-gif-23826808

[steel gyro]

https://youtu.be/AzGGZnQpF-s

[faint harness]

https://youtu.be/hOHYl3R1dwA

[quaint night]

What model you have used for this

[faint harness]

Custom merged model and custom trained Lora !

[quaint night]

How did you made this into video, what tool you have used

[faint harness]

Lots of attention to detail friend

https://github.com/AIManifest

I have a custom deforum notebook

Definitely for advanced use

[inner crest]

Mouahahah it works !!!

Ok time to mix that with Parseq and Control Net !

[crystal gust]

How did you swap face?

[inner crest]

https://youtube.com/watch?v=h6Sw1ltQKiA&feature=share9

The master nerdy rodent shows me the way

But it is not easy to install it

A lot of errors

[crystal gust]

I saw roop extension but I thought it was just for images

[inner crest]

And I wonder how many trojans I've installed with it

[inner crest]

The SD extension is for image

But you can install one for video

Without SD

[crystal gust]

Would it work on my Mac M1?

[inner crest]

I watched one tuto for mac

[crystal gust]

share the tutorial

[inner crest]

https://youtube.com/watch?v=W8xaFW-2UR8&feature=share9

I think it is this one

But just search

Roop + deepfake

I tried a lot of tuto. Only the one of nerdy works.

Problem was python

And I used anaconda

Nerdy is the best for weird and complex installation

[crystal gust]

Man with this, we won't need to train a model for every person. Its amazing!

[inner crest]

Yep

And the quality + consistency are great

[ebon sparrow]

https://youtu.be/xXd8LdzitTY

[inner crest]

I install the gpu version

it takes less than 5mn for making a video

I read there are many ways to improve quality

[toxic glen]

Very cool!!

Duuuuude! https://github.com/AIManifest/video-killed-the-radio-star

This looks tight af! Would you happen to have anything that was created with it?

I've actually messed around with lyric prompting quite a bit, it's super fun! Here's one with Aesop Rock

[faint harness]

Thanks friend! Check out https://github.com/dmarx. He's the original creator of it!

[toxic glen]

Yeah definitely, thanks. Do you have anything made with it that you can post?

[faint harness]

Not the newest inference, he created an entirely new system, but here’s what I have

https://youtu.be/jBYy_t6su4M

[toxic glen]

neat! Thanks for sharing

[dense zenith]

I'm kind of hooked on upscaling Deforum animations. It takes an absurd amount of time even with rent 3090 web gpu but imho it is worth it. These 3200 frames in this video took easily over 10 hours. I'm planning to do a 4k upscale with ControlNet tile sometimes soon 🙂

https://www.youtube.com/watch?v=lkV5_3KqOPc

[toxic glen]

Looks pretty cool though! What resolution are you rendering in initially?

[solar hearth]

Made with SD/Deforum/ControlNet and a little post in Resolve 18
https://www.youtube.com/shorts/9TVYbYTfUQo

[mossy ridge]

Sickkk just added u

[toxic glen]

daaaaaaammmmmmmnnnnnnnn that sht was clean

good stuff my guy

[gleaming tapir]

Can you explain the whole process

[dense zenith]

Thank you! I render almost all animations to half of full hd (960x540) and then upscale them by the factor of two

[toxic glen]

I just saw this online and dmed it to my buddy!

[last latch]

Eh much love bruv ❤️

[toxic glen]

that's one of your creations?

[toxic glen]

damn dude, that is next level cool

[inner crest]

And a special one for all of you who participated to the Peter Gabriel challenge

[dusky osprey]

https://youtu.be/eaokUSvyAzs

[last latch]

they're all my creations yeah

https://www.machinedelusions.com/

[hot dew]

A poetry video

[inner crest]

And Here's a short video about my reputation in ai art community:

[whole pelican]

why do you keep spamming the channel with seemingly random videos?

@inner crest

[inner crest]

because my goal is to post a video with deforum and show my workflow

[whole pelican]

your workflow is danicng wcith a tutu and coming out of a rhino's ass or what?

[inner crest]

last video I used green background and layers

you probably didnt notice it

because Im on SD server

I thought I was on deforum server

[whole pelican]

ah ok, i literally thought you are just posting unedited movie sccenes and i kept clicking away after seeing a few frames

[inner crest]

https://tenor.com/view/tony-stark-iron-man-rolls-eyes-rolling-eyes-i-dont-care-gif-22530187

[whole pelican]

like that yes

[inner crest]

Stable diffusion discord...

you want some waifu animation instead ?

[ebon sparrow]

https://twitter.com/iminescent/status/1673030022725095427?s=46&t=IqsmSp03G1H7FaBHr8WP1A

marc aaron (@iminescent)

#sakura sunday #stableanimation #sdxl #aianimation #latentspace #stabledifusin #stabilityai #dreamstudioai #diffusion #latent #animation #stopmotion #dreamstudio #surrealism #aiia #Aiart #latentspace #aiartcommunity #activeartcollective

▶ Play video

Twitter · Jun 25, 2023, 6:06 PM

[safe oriole]

i think they want lower quality videos that clearly look ai made 😄

how do you segment green screen videos thou seriously?

[inner crest]

https://tenor.com/view/whip-it-mediocrity-gif-23209054

[safe oriole]

it means you passed into making videos too good for human eye to quickly detect, which is fantastic

[inner crest]

I mean it is obvious Elon Musk played in all those movies

https://tenor.com/view/the-rock-smiling-hehe-gif-19148518

[safe oriole]

yea ole musky went back in time to star in a 1975 film 😄

thou actually wasnt the movie i thought it was , thought one flew over cookoos nest, my bad, got to get my eyes checked

[toxic glen]

Such awesome stuff! You're very talented and have a great style

[inner crest]

https://tenor.com/view/pass-saturday-night-live-smash-or-pass-next-not-interested-gif-25441284

https://tenor.com/view/gold-buzzer-heidi-klum-agt-agtgi-fs-americas-got-talent-gif-4295953

[inner crest]

Here's the workflow:

1- go to https://app.runwayml.com/video-tools/

you create a project and use somethings calls Magic tool + green screen

tuto here:

https://www.youtube.com/watch?v=5lltXmfQmbw

Obviously I dont use it only for background removing

but to isolate what I want to change

(After effect or Davinci do it also without having a green screen but for hundred of dollars )

Then you go to Davinci Resolve (free version works) and do this:

https://www.youtube.com/watch?v=AzpSHwG2P3Y&t=173s

or this (less efficient)

https://www.youtube.com/watch?v=PEXu5-i1TPk

So finally, you should have a video with your ai generated subject + original video

Right now, what I try to create is a mask transition:
https://www.youtube.com/watch?v=fKCv2J0f6gg&t=255s

It is annoying because, professional softwares can make this workflow in seconds

But they cost a lot

So I try to improvise and save money. Because I'm not a pro and it is just a hobby.

But everything I show are free. It just requires time and common sense.

[inner crest]

Thanks to mask, there is also a way to keep track of face in Davinci Resolve. But it is annoying and doest work as good as the website I give you

I wont lie, it may work for face swapping... Im not an expert in masking, I started yesterday

https://tenor.com/view/marvel-the-avengers-tony-stark-maria-hill-robert-downey-jr-gif-15685597

I know...

[safe oriole]

i am hoping the SD image tools to classify items into different classifies can be integrated with deforum to auto mask 'people' or 'whatever subject to auto mask' etc, being a bit hopeful maybe 😄

so at each frame it generates it would have to redo for the auto mask, maybe with auto roop too 😄

[inner crest]

there is something calls

https://github.com/continue-revolution/sd-webui-segment-anything

but I dont think it works with Deforum

more for images

[safe oriole]

yea thats what i was thinking of, if that could be integrated into deforum work flow 😄

where you specify what sort of segment to auto mask

[inner crest]

it would help a lot for sure

but basically, deforum would become a free competitor to Adobe After effect, Davinci resolve, Photoshop, etc...

because I think the core of these softs are the masking effect

it would be great

[safe oriole]

think about roop for instance, you could do an inverse roop swap 😄

[inner crest]

https://tenor.com/view/oh-amazing-amazing-fred-pye-nought-thats-perfect-thats-incredible-gif-26544998

[safe oriole]

controlnet video prepping, does anyone know how to clone 1 image 300x ?without copy paste copy paste and then bulk file rename and then merge into a video? (my current process to make my controlnet videos)

was browsing google on how to do in ffmpeg but no luck so far 😄 ill try some more later on tonight

[toxic glen]

So which part(s) are you wondering about

Bulk rename you can do on (certain versions of) windows

Stitching the images together for video can be done with ffmpeg, and right in a1111

To clone an image 300x, you can copy and paste it, copy and paste those 2, then those 4, then those 8, then those 16, then those 32... you get it

[solar hearth]

Your editing was so good they didn't even notice the changes LOL!!!

[inner crest]

Irfanview ?

[safe oriole]

oh how to turn 1 image into 300 without all of those steps 😄 on my ffmpeg searchs it suggested you can have 1 image generate for x seconds in a video, so it in effect clones that image directly into a video for you

[safe oriole]

thank you 🙂

[toxic glen]

deforum

Or copy and paste in windows explorer

[toxic glen]

if this can help you in any way, this part of my ongoing tutorial series is about using your own image(s) with deforum, which I guess could achieve the things you're looking to do
https://civitai.com/articles/606/dwh-2-animating-your-own-images-with-deforum

theoretically you should be able to use your image as an init image, and with a strength value of 1, every following frame would be a copy of the first

[safe oriole]

thank you 🙂 this is my current test video using 13 sketches my friend drew (used them to do as controlnet feeder video) https://www.youtube.com/watch?v=Sqzd1yvjNWc

[toxic glen]

pretty cool!

so is that what you're trying to accomplish?

Instead of like 300 identical frames

[safe oriole]

yea i hadnt thought about using 1 strength init image 😄 funnily i previously used init image default settings, and used the 'adjusted' output as the controlnet feeder video as it was close enough to the original 😄

[toxic glen]

I think youd want to lock the seed in as well

Not positive, never tried that before

i always want my animations to move/change 😂

from the tutorial, heres an image i made, then used it as an init image

there thats what i meant to post

[safe oriole]

i just found a fork of this for video and theres a twitter post someone removed advertising billboards from a video 😄 so i wonder if its in the wilds already

[inner crest]

On github ?

[safe oriole]

https://github.com/kadirnar/segment-anything-pip/releases/download/v0.2.2/metaseg_demo.gif

https://github.com/kadirnar/segment-anything-video

no idea if this is what they used for the billboards hiding example thou

[inner crest]

Thank you !

[inner crest]

It is exactly what I try to make right now

I try to generate the video with ControlNet right now

But I will test segment anything asap

But my final goal would be to have face swapping and be able to modify one or few elements in a realistic video

[safe oriole]

yea sadly i dont know which github they used to do the above

theres a lot of forks 😄

[inner crest]

Oh ok

But I know how to do it in Davinci Resolve now

So it is not a problem

And it takes me less than 5mn

But Im an ai purist

If I could do everything with stable diffusion, A1111 and extensions, I would do it.

[toxic glen]

Yeah the creator (Fill) just posted that above

It reminds me so much of Snackys work

[heady lark]

any guides for faceswapping videos with roop in A1111? is it possible?

trying with sd-cn animation but the face im trying to swap isn't really sitting on the face as the face moves, do i need to use controlnet?

[safe oriole]

oh good thinking with sd-cn , think people roop outside of a1111 for videos but shrugs i havent tried either

[stable hawk]

checking out the updates with Infinite Zoom and SD-CN since I haven't used either in a hot minute lol. First part is Infinute Zoom, then SD-CN takes the last frame of that for it's video. All settings default to see baseline quality.

[inner crest]

it is not used in A111 for videos

you need it to install it beside

https://www.youtube.com/watch?v=h6Sw1ltQKiA

Nerdy rodent made the best installation tuto I've watched

Problem is that thing works only on Python 3.10

so you need to install anaconda and follow the steps of master nerdy rodent

Also, in order to use gpu

you install the depedencies:
pip uninstall onnxruntime onnxruntime-gpu
pip install onnxruntime-gpu

Then use this command to launch the app

python run.py --execution-provider cuda --execution-threads 1

you may change 1 for 2 or 3 or 4, etc...

"--execution-threads" allows to not overwhelm your vram

useful when you have a small gpu

if you have a 3080ti at 24gb, you dont care

https://github.com/s0md3v/roop/wiki/4.-Troubleshooting

[heady lark]

i tried this when it came out but i got a warning about some trojan in one of the python installer files so i just abandoned using it

[heady lark]

I tried it using the SD-CN animation and it did kind of work

i do believe it can be used

[inner crest]

the warning is because of the model

inswapper_128.onnx

[heady lark]

but ofc it's nothing like the standalone one

[heady lark]

really/

[inner crest]

that things tries to download it on civitai

I got the same warning

and I just manually install it

[heady lark]

from your anti virus/

?

[inner crest]

yep

[heady lark]

i dont recall that being the name but i shat myself when it appeared

was like, im not risking it lol

surely there should be a way to acheive the same effect as the standalone version, in A1111 ?

[inner crest]

I guess batch img2img + CN

but roop is really fast

like in less than 5mn it makes you a full video

But the thing is funny

I try to create a Musk like dr manhattan

but roop dont recognise the picture above as a face

I mean Musk is a superhuman but he is still partially human

I think it is because of the blue

[heady lark]

really? what gpu are you using?

[inner crest]

roop ! 😄

[heady lark]

what, in cpu mode?

[inner crest]

gpu mode

[inner crest]

Ha ha ha, it finally works !!!

this video shows how tech industry works...

and I start to understand what roop needs to track the face correctly and how to fix it...

[heady lark]

you did this in sd?

[heady lark]

right so i mean you say you did a video in 5 mins, what gpu are you using and how long was the video?

[heady lark]

how long did that take to make? zuck at the end lol

[inner crest]

DaVinci Resolve

[inner crest]

46seconds and 8gb nvidia 2080

[inner crest]

it is less than 5mn to generate. Editing in Davinci is like 20mn.

[heady lark]

oh so that is just standalone roop and davinci resolve?

[inner crest]

yes

I try to add controlnet and deforum animation in video now

it takes time

[inner crest]

https://www.youtube.com/watch?v=8SWon8erHZw

that thing is going viral

[heady lark]

thank god we can talk about roop here, I got banned yesterday on the unstable diffusion discord for asking if anyone had used roop in A1111 to deep fake (ok,i should probably have used the term face swap because im not trying to do anything malicious with it and deep fake has those negative connotations but still.. funny coming from them)

[heady lark]

did you have any luck? im guessing it's nowhere near as good as the standalone app? i tred with sd-cn animation and it worked but typically messy

[whole pelican]

It would've been against the rules to talk about it no matter the term. They are a porn discord so deep fakes being against the rules makes more sense. Just read the rules of discords you join

(also tbf at least until recently, and as far as I can see maybe still currently the deep fakes with SD are at best as good as 2020 deepfacelab)

[heady lark]

yeah, fair play, their rules but a bit silly to ban without warning, i'd been a good member, i've appealed hopefully they see sense I'd been in there since almost the start, tbf roop didn't exist back then so i had no reason to ask about it. anyway. I actually asked there first I think as I thought THIS place would be more hostile lol, ironic

[inner crest]

using deepfake for porn when you can make this:

https://tenor.com/view/human-brains-are-primitve-and-occasionally-malfunction-humour-fact-about-human-truth-about-human-wisecrack-gif-14678502

[inner crest]

I just try to use mask + tracking in Davinci Resolve right now

I will go back to CN + SD as soon as I learn how that works

[elder sable]

{0: 'masterpiece:1.2, A shirtless man stands in the center of a dimly lit gym, surrounded by boxing equipment, cinematic:1.2, 8k ultrafine detail, caption: a shirtless man wearing boxing gloves in a gym', 44: 'masterpiece:1.2, The man begins to warm up, stretching his muscles and shadowboxing, preparing for his training session, cinematic:1.1, 8k ultrafine detail', 88: 'masterpiece:1.2, The man steps into the ring, his gloves laced up tight, ready to face his opponent, cinematic:1.2, 8k ultrafine detail', 132: 'masterpiece:1.2, The first round begins, the man and his opponent exchanging blows, sweat flying, cinematic:1.3, 8k ultrafine detail', 176: 'masterpiece:1.2, The man takes a hit, stumbling back, but quickly regains his footing, determination in his eyes, cinematic:1.2, 8k ultrafine detail', 220: 'masterpiece:1.2, The second round begins, the man unleashing a flurry of punches, his opponent struggling to keep up, cinematic:1.3, 8k ultrafine detail', 264: 'masterpiece:1.2, The man lands a powerful blow, sending his opponent to the mat, the crowd erupting in cheers, cinematic:1.4, 8k ultrafine detail', 308: 'masterpiece:1.2, The man raises his arms in victory, his gloves held high, cinematic:1.3, 8k ultrafine detail', 352: 'masterpiece:1.2, The man exits the gym, stepping out into the misty mountains, his breath visible in the cold air, cinematic:1.4, 8k ultrafine detail'}

Whats missing why animation is not syncing with prompt?Any plumbing work am I missing,lmk?

also the color got faded whichI was nto expecting

[whole pelican]

the prompts are just seperate image prompts, not actual description of an animation

think of what you get when you do just an image prompt for 'The man begins to warm up, stretching his muscles and shadowboxing, preparing for his training session'

[elder sable]

yeah I felt so, what can bring the convergence?

[whole pelican]

the kind of animation you want to do is really hard to do straight from text, there are some txt2vid stuff like SD-CN you can try though https://github.com/volotat/SD-CN-Animation

[elder sable]

I am providing one seed image too

[elder sable]

let me check may be this comes handy

[whole pelican]

if you want it to stay more like the seed image you'd want to use Controlnet and adjust how important the seed image is for a start

but in general for more complex stuff you want to find a workflow video/comment (which would also be typically more complex than just vanilla deforum use) and go from there

[elder sable]

do deforum offer those api capaibility?

[hot dew]

Any hindi music fans here?

[dusky osprey]

https://youtu.be/zOpqCKfL1NA

[mossy ridge]

https://www.youtube.com/watch?v=iEf7O7K_JTg&ab_channel=CalmCadence

[elder sable]

I used openai api to generate this prompt, my input were [action] [num_frames][scene:balkan or desert]

[rancid jewel]

Please rate my videos made using stable diffusion--https://www.youtube.com/channel/UCAoj6zVyGHIpVo_U8pbSWsg

[inner crest]

Musk vs Zuckerberg. The fight of the Century !!!

[ebon sparrow]

https://youtu.be/txRW0P3qYmg

[safe oriole]

i hear this is going well , congrats 🙂

[inner crest]

Thank you

https://tenor.com/view/kind-of-a-big-deal-anchorman-ron-burgundy-smile-gif-16393820

[inner crest]

They removed my video after bringing 200k views to the sub, it is ridiculous 😂

I made more views than the last show of Peter Gabriel with that video !

https://tenor.com/view/wake-the-hell-up-the-black-mastadonte-wake-up-get-up-rise-and-shine-gif-26134087

[toxic glen]

Yeah man Reddit views are crazy

[heady lark]

Does anyone have tips for prompting with text2video?

most of the results i get are pretty bad, is there a specific way to prompt for 'actions' ?

(as it's a moving image an not just a picture?)

[rancid jewel]

https://youtu.be/A1pk78NSv1M

[ebon sparrow]

Control net m2m

The original video is free polygon

[distant lion]

One of the best fighting scenes in movie history 👌🏻😄

https://youtube.com/shorts/rx47lDye2O0?feature=share3 I tested some audioreactive stuff in Touchdesigner combined with SD. Quality is bad but you get the idea, super cool 🥳

[willow quarry]

do you guys have any recommendations for sprite making websites

[hot dew]

Does anyone have an idea how this is done? Is it wonder dynamics?
https://www.instagram.com/reel/CsQKWd3AXnY/?igshid=MTc4MmM1YmI2Ng==

Instagram

[willow quarry]

ai made my own freddy fazbear
Image
Image
Image

which do you like

[mossy ridge]

guys can someone name the ai used for this animation?
https://www.youtube.com/watch?v=JWdkHdZ5ILk
ive seen couple animations like this but never found out

[nova quartz]

https://www.youtube.com/watch?v=bFHmmyL40L0&t=68s&ab_channel=AIMiku

[distant lion]

Looks like Deforum extension for A1111 WebUI

[toxic swan]

geeks fr

[rancid jewel]

Good time of day. Please rate my videos stylized in stable diffusion. I will be glad for any comments and criticism-https://youtu.be/ZG8dbNx43OY

[sharp depot]

https://youtu.be/MEN3FAOxed4

[fallen girder]

"Chaos Protocol" An AI Generated Short Film made with Deforum 3d mode, 100% Text2img without source video or ControlNet, animated talking faces made with prompts
https://youtu.be/Z3gq--NwUx8

[inner crest]

did you use parseq in order to open/close the mouth

https://tenor.com/view/frickin-awesome-amazing-gif-20449310

[fallen girder]

it's just the classic 3d mode, i haven't tried parseq yet, too many things to learn 😀 , i used this:

[inner crest]

oh damn, amazing

I can see how long and hard it must have been

above all without parseq

https://tenor.com/view/oscars-standing-ovation-clap-clapping-applause-gif-20292436

[fallen girder]

Thank you very much! ❤️‍🔥 it took me about 2 weeks, a lot of work

[obtuse jungle]

How do yall get such consistent results, im still learning controlnet but i haven't been able to render anything as good as some of these

[ebon sparrow]

The model is using "Flat-2D Animerge" and the prompts are also using this site as a reference.
ControlNet (v1.1) uses the "lineart" model.
https://civitai.com/models/35960/flat-2d-animerge

[kindred finch]

An Interdimensional Travel Show made with ai https://youtu.be/GD_d-MVn84M

[ebon sparrow]

ControlNet m2m in SHIBUYA

[void girder]

https://www.youtube.com/watch?v=-OkjBB1WJTY I tried to make a normal commercial and the ai went wild hahaha

[ruby nova]

goddamn

[void girder]

😂

[livid wasp]

https://youtube.com/shorts/oO52UdG-ud4

[toxic glen]

Looks to me like 3D animation with Deforum, using their own "init image".
I have a tutorial on how to do this on civitai:
https://civitai.com/articles/606/dwh-2-animating-your-own-images-with-deforum

[toxic glen]

It depends on what type of consistency you mean specifically I guess. As far as just creating clean looking animations, that can be achieved with nothing more than Deforum and a little finesse (control net sure doesn't hurt)
For vid2vid, or animating over video, warp fusion is where the best "consistency" is, as far as I've seen at this point.

[obtuse jungle]

Is warpfusion on a google colab notebook somewhere? I tried finding one hosted onngoogle colab but i didn't see one

I made this one with SD-CN-Animation

[toxic glen]

It's only available through the patreon

[obtuse jungle]

Do you have the official link?

To that

[toxic glen]

https://www.patreon.com/posts/stable-v0-5-12-71681839

I feel like I might as well post this any time I even talk about Warp Fusion at all

[obtuse jungle]

@toxic glen thank you 😁

[jagged vapor]

Decided to try and make a music video for some music I made a few years back using Deforum. Here is my first attempt at using Deforum!

https://youtu.be/oRKiUdnhiu4

[normal latch]

How did you do that , explain me , it will be great of you ?

[near pagoda]

Welp. I'm using SD_CN and my GPU sounds angry.

I do like that it shows a ETA

[ebon sparrow]

Thanks for the reaction!
The video is processed by ControlNet as a still image one by one.

The original video is below
https://media.discordapp.net/attachments/1069951768164847687/1125816904976703558/514976812.mp4

[limber dagger]

Hi Guys, I'm a beginner ...I was trying to use Control Net. Could some help me out with more clear face in the video. How to get more clear and precise face?

[fallen girder]

"Danse Macabre" | Zeroscope Text2video + Deforum / ControlNet
https://youtu.be/BGnAgjh1wW4

[fringe carbon]

Any of y'all good with Zeroscape? All I've managed to do is add some barely noticable car movement. I need the camera to move around and stuff, how is that done?

https://i.imgur.com/DIz7005.mp4

[unique palm]

i just got a crazy idea you could take a screencast video of cities skylines and run it thru stablediffusion

the camera motion isn't smooth enough though

[fringe carbon]

🤔

I've added the scenes with birds to the vid above. Doesn't look too bad I guess.

[void girder]

https://www.youtube.com/watch?v=fFfEExQS1mk

[solar hearth]

https://vm.tiktok.com/ZM29Chm1A/
People are really enjoying this one.

[untold pulsar]

https://youtu.be/GAGnQuisyxg
Bringing the Ramayana Epic to Life with SD & Depth 3D

[void girder]

https://www.youtube.com/watch?v=0N6Sz2JXtls

[jovial magnet]

Charlize Theron riding a tiger

[proper dagger]

That looks amazing what is your workflow?

[dense zenith]

Melting faces with Deforum
https://youtube.com/shorts/X1GvgVPxjHw?feature=share

[light grove]

I can post it in reddit

[proper dagger]

I would definitely be interested

[light grove]

workflow is there : https://www.reddit.com/r/StableDiffusion/comments/14v2owl/toy_story_if_made_in_japan_accidental/

[vocal saddle]

The purple artifacts on shoulder, is because you're not using vae

[fallen girder]

Zeroscope v2 XL Text2video + Deforum / ControlNet
https://youtu.be/s7xJbG7lo_I

[coral violet]

hyper realistic animated portrait of a hindu god lord Krishna playing flute, insanely beautiful, having peacock feather in hair, digital ui, by Guy Denning, Johannes Itten, Russ Mills, glitch art, hacking effects, glitch effects, digital tech effects, chromatic, color blocking!, oil on canvas, concept art, abstract,shoe with snake as a lace style:Fantasy Art aspect:9:16

[fallow sierra]

This looks amazing. Can you pls detail out your process a bit more.

#🎥｜animation
Made with deforum and after effects. Music made on logicX with Sub 37. https://www.instagram.com/reel/CuT9vDLgVNr/?igshid=MzRlODBiNWFlZA==

Instagram

#🎥｜animation

[pure zinc]

https://twitter.com/eddy_phast/status/1678881878906994688

PhastEddy (@eddy_phast)

@ClaireSilver12 ✌️

▶ Play video

Twitter · Jul 11, 2023, 9:40 PM

[wraith flame]

what was your controlNet models and preprocessors?. Did you use image2image batch?

[stray yarrow]

https://youtu.be/oPBhjNp5dI8

[gilded bane]

Made with Kaiber

[vernal folio]

Deforum using colab

[near pagoda]

Kaiber takes things to a whole new level.

[raw scarab]

https://youtube.com/shorts/OHdGhc6b49I?feature=share3

[gritty rain]

Deforum (mixing multiple ControlNets with Hybrid Motion). Upscaled and interpolated in Topaz. Music by me.

[frank kiln]

just dropped a music video I made using stable diffusion and inpainting to make a claymation style https://www.youtube.com/watch?v=ZzW3Gvjw9YQ

[pulsar wyvern]

https://youtu.be/avIcRrQNBm4

[pulsar wyvern]

https://youtube.com/shorts/CEb2H6EemSQ?feature=share

[faint rivet]

https://www.youtube.com/watch?v=cq_D3e3oZWk

[undone badge]

dafuq i just saw

[gritty rain]

Chillin', Deforum style

[void girder]

https://www.youtube.com/watch?v=sWUzP7FB-o8

[wispy oasis]

by stable animation sdk

[mental escarp]

Tree leaves

[pulsar wyvern]

https://youtube.com/shorts/Y14uDfZ3vMo?feature=share

Did the quality should be improved

[mossy ridge]

https://m.youtube.com/watch?v=DMH5P_h93qs&t=400s

[void girder]

https://www.youtube.com/watch?v=rSDGbxEcseI

[steel pilot]

hello

[fallen girder]

"Street Runner" | Zeroscope XL Text2video [4K]
https://youtu.be/gLdFf-4a51Y

[steel pilot]

!create "An animation of a cat transforming into a dog over time" 3-4

[pulsar wyvern]

https://youtube.com/shorts/ODernuI-BiM?feature=share

[mossy ridge]

Hi nice video. @wispy oasis

[proud cave]

"Text-to-video/gif" AnimateDiff

https://github.com/guoyww/AnimateDiff

[serene jay]

has anyone been able to run this on anything below a 3090/4090?

[proud cave]

no idea i used google colab

[fallen girder]

Zeroscope XL + Deforum / ControlNet
https://youtu.be/3fVgBCPsHKg

[bronze tiger]

Is it available in i2i or controlnet now?

[proud cave]

as far as I know.. Nope

[bronze tiger]

it is only mentioned in their paper but nothing more than that

[main pasture]

hi

What are the models that use stable to generate videos? Can someone guide me?

I have installed 1.5, but the effect of the generated video is not particularly good, is there any other model that can generate video？

[serene jay]

what kind of video are you trying to make?

there are quite a few tools from kaiber to deforum to zeroscope and more

[main pasture]

I want the things in the generated video to be more realistic, not animation

On the other hand, I want to make the generated video more clear

[serene jay]

maybe you should look into zeroscope and/or runway

[main pasture]

but they are not provide the api

I need api

[serene jay]

are you trying to make a video or a service to make videos?

[main pasture]

yes

[serene jay]

if it's the latter you should probably look into zeroscope

[main pasture]

I want to provide the servie fou many company

[main pasture]

stable Are there other models that generate videos, except for the v series?

[serene jay]

text to video you have modelscope and zeroscope

[main pasture]

can I use this model to general the video?

[ripe shuttle]

Here is my newest stable diffusion infinite zoom video

https://www.youtube.com/watch?v=qow6nHJb9o8

[livid wasp]

https://youtube.com/shorts/Gn9ym-ay1-Q

[formal perch]

A Bird

[kindred monolith]

Hey there, trying out deforum for the first time, followed a few tutorials on yt, I finally can get pretty consistant animation, yet this "doubling" happens, tried messing with the parameters a bit but it doesn't seem to really go away, I still don't understand how the tool works fully so I'm just shooting in the dark. Maybe someone has any ideas why this is haappening?

I think optical flow was causing this! Still have to do some tests, but on a short clip seemed to fix the issue:)

[serene jay]

it looks like weird interpolation, optical flow depending on your cadence and method changes a lot and can give weird artifacts

[kindred monolith]

Ah I see! I don't really understand cadence, it says "# of in-between frames that will not be directly diffused" what does "diffused" mean in this context?

[serene jay]

cadence only makes inference in the specified number of frames.. every 2, every 4 or whatever number you put there... it will infer, then interpolate until the next inference

if you use higher cadence you can have better temporal coherence

[serene jay]

sorry i mean diffuse, not inference

[kindred monolith]

aha, that makes sense!

Could you ELI5 what does diffuse mean in this context?

I'm still very fresh so trying to get with the terminology:

[serene jay]

inference happens on a prompt. diffusing is what's generating the image from noise to gen.

you can diffuse in every frame, it will look really sharp but very "flickery"

or you can give some cadence and interpolate

with cadence you have different methods for interpolation

This is very helpful> https://rentry.org/AnimAnon-Deforum

[kindred monolith]

Ooooh! That ties some loose ends for me, thanks so much!

[kindred monolith]

amazing, thanks so much!

Would you have any other material you'd recommend to read?

[serene jay]

there is a very complete site about animation tools curated by one of the deforum devs> https://pharmapsychotic.com/tools.html

you can also check our animation competition channel, people usually post cool stuff there too> https://discord.com/channels/1002292111942635562/1097876660101066854

[kindred monolith]

Thanks so so much!

[serene jay]

Just a friendly reminder, we need everyone to go and vote on the bot channels, please select what image you like best: "A" or "B", this is for helping improve and finetune what will be the gold ⭐ SDXL 1.0 model.

[gritty rain]

Created in Deforum https://youtu.be/XF7ZBDOwqAQ

[inner crest]

Musk visiting the worst toilet in Scotland.

[stray tree]

Stable Difussion + Audio Reaction
https://www.youtube.com/watch?v=U9srIgTISTw

[woven cipher]

I have been making cover songs in Minecraft for a couple years now, but recently started including animations related to the song off of a single image with Stable Diffusion, ControlNet and Deforum. I like how this Touhou cover 'Solar Sect of Mystic Wisdom ~ Nuclear Fusion' turned out:

https://www.youtube.com/watch?v=VqZFwr1EcN4

[misty dagger]

SDXL

[light grove]

Weeb dream

[mental pecan]

https://vimeo.com/831329933

[delicate briar]

the first part of the time traveler

[void girder]

https://www.youtube.com/watch?v=46NAQnHvBt0

[vapid skiff]

FAQ: Is Stable Diffusion open sourced?

Yes! Even though we offer competitive rates for the computational cost of generating images on https://beta.dreamstudio.ai, you are also able to run the various open sourced versions of Stable Diffusion using your own hardware without any additional costs! Take a look at #1080946152318443610 to see how to get started!

[proper dagger]

How is this one done?

[clear horizon]

runway gen2

[proper dagger]

Oh ok

[dense zenith]

This Hybrid Deforum got pretty interesting 👽: https://youtube.com/shorts/Fw3OHnLaw5w

[solar flax]

Working on a gif plugin for EasyDiffusion 🙂

[solar flax]

Sadly didn't get a reply in the SDXL stream, but anyone here tried using SLERP to create new composite words ?

Here is an example of using to make a cat that slowly turns into a frog.
https://gfycat.com/flickeringdiscreteasianlion

[lusty star]

should apply some rife to that

sdxl^

[solar flax]

It already takes me half an hour to make a 50 frames gif with SD2.1

I can't wait to start messing with the code for SDXL, but sadly each time I'll want to see the results of the changes I know that I'll have to leave it running for at least an hour if not more...

[void girder]

skibidi toilet AI 😬 https://www.youtube.com/watch?v=uB5FAlS1V4k

[summer mural]

My 9 lives: https://youtu.be/JKzPh3l9tlg

[dense crater]

https://www.youtube.com/watch?v=FImW9yYPo3Q

[narrow walrus]

https://youtu.be/p__zeyPdrWc

[livid wasp]

first SDXL video 😈
https://youtube.com/shorts/U3-LmYx5IvY

[cunning flare]

Curious young boy named Alex, standing in his dimly lit attic, holding an old, dusty umbrella in his hands. The attic is filled with various forgotten items, and rays of light from a small window illuminate the magical umbrella. #🎥 | animations

Curious young boy named Alex, standing in his dimly lit attic, holding an old, dusty umbrella in his hands. The attic is filled with various forgotten items, and rays of light from a small window illuminate the magical umbrella. #🎥 | animations

[pulsar wyvern]

https://youtube.com/shorts/okfnkGhgDPU?feature=share

[dense crater]

https://www.youtube.com/watch?v=nIGdDrF15XU

[real marsh]

adding runaway and way2lip for a compelling result
https://www.youtube.com/shorts/LOPkaQcgQDI

[winged tusk]

https://user-images.githubusercontent.com/7041726/255722793-a6513aff-0b5d-493d-b34f-cc54c811b69e.gif

Fake deforum effect, example available here: https://github.com/melMass/comfy_mtb/wiki/Examples (requires my node pack)

[winged tusk]

blue car -> yellow car
https://user-images.githubusercontent.com/7041726/255719495-c15644ba-2435-4ed4-afa9-e7620f3cf41c.gif

[void girder]

https://www.youtube.com/watch?v=hc21g7xQv2Y

[vast nebula]

https://youtu.be/8bdRku4edyU

https://youtu.be/sr_1eoDKDsw

https://youtu.be/80hxK3rom4M

https://youtu.be/qqshgpka1GY

https://youtu.be/yvZkG8bPFos

https://youtu.be/X-XLuIXPZp4

[winged tusk]

@vast nebula What is this?!

[wraith flame]

made with SDXL and PikaLabs

[digital topaz]

https://cdn.discordapp.com/attachments/1025266140445933648/1135090606931181698/Screenshot_20230728-202048.png. Make him a statue

[digital topaz]

Hey how you do this? Help me out by step by step!

[winged tusk]

You forgot please 😅

There is a sample workflow here: https://github.com/melMass/comfy_mtb/wiki/Examples

[lusty star]

https://www.youtube.com/shorts/bm-RnUaeyCk?vq=hd720

[rough vessel]

https://www.youtube.com/watch?v=7YB2DcjYmDc

[vapid skiff]

@near badger

FAQ: How do I generate images? Is there a bot on the server?

Currently, there is a public bot on the server that generates images available as a research beta for SDXL, you can find the current status of the bot in #1047610792226340935. There are plenty of ways to use Stable Diffusion such as the official https://dreamstudio.ai/ website or by running Stable Diffusion locally using your own hardware - check out #1080946152318443610⁠ for more details! You can also stop by #1025467151206854736 for any issues you experience while using DreamStudio or #🤝｜tech-support for any problems you encounter while installing it locally!

[solemn oxide]

how much gpu mem is needed for sdxl base 1.0?

[fallen girder]

Dissent 異議 | AI Generated Short Film (Zeroscope v2 Text2video)
https://youtu.be/TEm4kkVSgMQ

[mossy ridge]

sdxl + DPM++ sde 2m karras + parseq

how could I fix the eyes on this one next time?

[icy lava]

does anyone have an idea on how runway is implementing their image to video functionality?

[void girder]

https://www.youtube.com/watch?v=Oz3FPaOgHA4

[real marsh]

what about a talking photo?
https://youtube.com/shorts/LOPkaQcgQDI?feature=share

[worn gust]

how to create link please?

[worn gust]

how to create Link please?

[mossy ridge]

start by getting sum youtube tutorials on what do you want to create, there are many branches and styles, and ways to achieve stuff

[jade prairie]

有人知道如何实现图生图吗？

[ornate sage]

我建议您用英语提问，因为这样会有更多人参与。

翻译：DeepL

[vast nebula]

https://youtu.be/_Ozk-_maSFQ

[toxic glen]

If you want to get started with Deforum, I've started a tutorial series on Civitat that will teach you the basics and get you started
https://civitai.com/articles/554/dwh-1-creating-your-first-animation-with-deforum-in-1-click

[radiant scroll]

Images: "Street art multiple happy connected brains in a classroom"

[fallen girder]

Rick Batyr Ft. Dontmesswithjuan - BOB (Official Music Video)
https://youtu.be/6Pm0RSmwxj8

[toxic glen]

Wonderful video! Really great job

[livid wasp]

https://youtube.com/shorts/lHeiXN97IMI

[fallen girder]

Thanks a lot!

[crude solar]

Hi all,
if you need a good wav2lip for your video you can try this extension in Automatic1111:
https://github.com/numz/sd-wav2lip-uhq

[solemn flame]

Hi, II am a beginner, plus I have a 2060... I suppose I cannot do too much... I have tried prompt travel... any suggestion for low vram sd animation tools? sad talker works fine but at low res... Should I try Deforum? And which model? thank you

[empty delta]

hello

2060 6GB or 12GB? either way https://github.com/neggles/animatediff-cli should let you generate infinite length gifs/frame sequences at 512x512 with an 8-12 frame context, just barely

[jaunty egret]

I spent the last 3 weeks working on this animation and it is finally done!
A full animated music video using Stable Diffusion, Dreambooth, and ControlNet

https://youtu.be/FoBMEB5Dock

[empty delta]

https://giphy.com/gifs/lGRwq4U3pKa9I6oHj9

need slightly different params to stop the jumpiness but

[jagged vapor]

https://youtu.be/FraHy9Sngfg

My second attempt at a music video using Deforum!

[topaz snow]

first time trying animated stuff!

i like how it turned out but i feel like theres too much changing in each frame is there anyways i can make it blend more with frames? if that makes sense

[serene jay]

is it a video init running on deforum?

if it is controlnets you can try using a higher cadence, for video init you can't but you can try playing with the noise scheduler and values

[topaz snow]

idk how to use deforum still 😅 i did that on control net yes

whats a higher cadence

[serene jay]

in deforum, diffusion cadence skips the number of frames you specify there

there is a cadence parameter, if you set it for instance at "4", it will only diffuse every 4 frames

so the final look will be less jumpy, images will change only every 4 frames instead of each one

fyi the strenght parameter and the noise parameter can also help to get a more temporal coherent look

[lyric niche]

Really cool!

[crude solar]

thanks 😊

[topaz snow]

alright, thanks so much 😭 i tried this settings and it does look way more smooth and less jumpy like you said but deforum is still a bit confusing for me to use lmao, ill mess around with more settings and share the results later!

[raven fulcrum]

Neon City

[topaz snow]

is it possible to convert a normal video to ai animation using deforum? like you can do with controlnet where you input a video and turn that into ai animation

[solemn flame]

6 GB... thanks for the advice 🙂

[faint rivet]

https://youtu.be/Tgqm8vxVFCA

[topaz snow]

i made this with deforum (controlnet NOT enabled) some of the frames are reallt deformed, i had a bunch of negative prompts for it and still came out like that

also i tried the cadance settings and it still kind of feels a little too jump 😅 how can i improve this

oh and last thing im using meinamix for checkpoint model which idk if its meant for animating so any good model for animation would be appreciated, ty

[serene jay]

for a more coherent look you can chain a couple controlnets with different processors. you can also try looking at what warpfusion does, which could be what you are looking for

[raven fulcrum]

Incredible Hulk

[rough vessel]

Asuna
https://www.youtube.com/watch?v=fEBPhU3z6DE

[raven fulcrum]

just....wow, generated from a regular AI 2D image

[wraith flame]

Made with SDXL, PikaLabs and After Effects for camera movement.

[sage fiber]

made with sdxl1.0 base images and piped thru pikalabs

[dense crater]

https://www.youtube.com/watch?v=JFnDNv-tD74

[topaz snow]

i looked up warpfusion and yes! its what im trying to do but the app seems paid, is it possible to get that kind of result on deforum or controlnet?

[cedar oxide]

https://youtube.com/shorts/2jKntep4yKs

[serene jay]

originally it was only available via patron but now it's public, let me see if I can get you the github... still if you find it useful it's always nice to support the devs : )

[topaz snow]

Thanks! I appreciate it. I 100% don't mind supporting the Dev's when I'm able to afford it 😭

[serene jay]

https://github.com/Sxela/WarpFusion

[topaz snow]

🥲 thanks so much, you're the goat

[serene jay]

you're welcomed... keep posting your experiments and advances

[topaz snow]

Absolutely

[empty delta]

i believe (under linux at least) 6GB juuuuuuust works with an 8-frame context

just

[somber lintel]

hey guys. I made a little tool to help write timeline animations in deforum. i'm looking for people to try it and give me feedback
https://deforum-helper.vercel.app/

[solemn flame]

yes, I am on Arch Linux infact... I have noticed most of you use comfyui... I am on WebUI by AUTOMATIC1111

[solemn flame]

I have another question... I have just installed Deforum extension just to try it out... What is the difference between deforum and Prompt Travel ?(except for the fact that PT dosn't request JSON format)

[raven fulcrum]

Awesome!

[topaz snow]

not perfect yet (obv) but im getting close i feel like lol

[orchid palm]

has anyone used posemy.art before? If i press "g" to move a joint, it just defaults to the entire character and not a specific joint, and im wondering if im doing something wrong or if this is a limitation of the free version?

[orchid palm]

in the tutorial im following the guy sets preprocessor to none and model to openpose, but my model parameter has an empty dropdown menu and i can't select anything.

i can only set the preprocessor to openpose

[rough vessel]

cause you have selected openpose on control type

[orchid palm]

it turns out i had to download the openpost pth file and stick it in models first.

[somber lintel]

Demo of my deforum timeline helper app. 😁 try it and give me feedback so i could make it better https://deforum-helper.vercel.app/

[rough vessel]

you could try to add some "presets" menu with some premade settings like 360 char turnaround or spinning zoom?

[somber lintel]

Great idea. I'll add it to my to do list 🤘

[rough vessel]

🙇‍♂️

[vast nebula]

https://youtu.be/Mfj1a4NNNAM

[topaz snow]

great animation bro

[topaz snow]

after hours of messing around i feel like im fnally getting somewhere lol

im using mask video now and it doesnt animate the bg so if anyone knows how to fix that, id appreciate the tips 😭

[pulsar heath]

Can anyone give me a starting point for making a few framed animation of a plant growing?

Everything I find online uses an existing animation first

[rough vessel]

try with animatediff

[pulsar heath]

Do you have any guides?

[rough vessel]

https://www.youtube.com/watch?v=rdnOhM8L8nE

[pulsar heath]

Brilliant, thx

[digital topaz]

Which tool is you used for this?

[topaz snow]

Deforum with control net!

[digital topaz]

Can we do in this bot, stable foundation?

[topaz snow]

absolutely

[serene jay]

https://tenor.com/view/confused-look-dog-gif-25616231

[serene jay]

on the bots in this server you can only generate static images using sdxl, for animation there are multiple tools and apps but the server bot does not animate

[topaz snow]

oh wait my bad, i thought they meant if those tools are available on stable diffusion 😭 sorry

[serene jay]

no worries!

[topaz snow]

damn what da heck, thats beautiful

[serene jay]

thanks!

[topaz snow]

[void girder]

https://www.youtube.com/watch?v=Rizxr8A5vMs

[fallen girder]

"Space Traveler" | Deforum Animation (SDXL 1.0)
https://youtu.be/9mDyg3seRPg

[sage fiber]

i have rewritten animatediff

look at this perfect background. also new motion model

[fallen girder]

SDXL Deforum Hybrid motion, optical flow RAFT

[somber lintel]

I made a little overview guide for the timeline editor tool. Sorry about my poor English in advance. Hope it will help
https://youtu.be/sskloFZXCLM
Deforum timeline helper - https://deforum-helper.vercel.app/

[stray tree]

My first animation with gen2:
https://www.youtube.com/watch?v=oSNbNKu3sMw

[jagged vapor]

https://youtu.be/Xw879oXdtwg

[vast nebula]

https://youtu.be/5iZ98A5OgQI

https://youtu.be/uWXzwXUVACI

[oak urchin]

Deforum + Parseq ; and sdxl
Fullscreen required - 768*1216
Test (371Mo)

[hybrid musk]

any tutorial on how to create easy animations with ai?

[oak urchin]

And here is the interpolated version

[oak urchin]

Hum, there are a lot. Depends what you're looking for.

Like, your very first one ? More text2vid, text2img or more img2img

You can install Deforum from the extension tab (a1111 webui) and press generate (default deforum settings), for a very first one.

Ask, would be happy to help to the basics

Here are the Deforum (+ Parseq) settings for the video above (file to load in deforum)

[hybrid musk]

I would be curious on text2vid, never did anything on ai animation, only images, so yea
I have stable diffusion installed locally, idk if this has to do

[oak urchin]

Okay, then, i would just insist on the difference between pure text2vid, and the "text2vid" (in fact text2img, but assembled into a vid)
I never use a video as input, cause i see this as "a cheat". I'm interested in pure text2img generation.
You should give a try to Deforum then. Install from the extensions tab, restart and see

You set a prompt (Or several prompts), and put motion settings. Can go way further of course

I made a chatgpt prompter for deforum months ago. You can find some examples in the gallery, with settings.

https://civitai.com/models/59785/deforum-chatgpt-prompter

The video above use something else, called Parseq. It allow to manage your settings, but I use it to synchronize with a music.

Run the default settings, and see what's going on, no need to look at every params, focus on the not empty/null ones

Default settings video (120 frames):

Once, i made an 11 min long video, using this prompter.
Downgraded the quality, and not full, but here is the travel through Parallel Worlds. (10 000 frames)

You can join the Deforum discord

[rough vessel]

do you have the settings file for this video?

[oak urchin]

lol yeah, somewhere

[rough vessel]

🙏

[oak urchin]

I think it's this one

Made with the prompter

[rough vessel]

ty sir 🙇‍♂️

[oak urchin]

Here are some material (also this one)

some stories as prompts and a few motions, this :

you can cut or extend them, it's a pattern

Like a plane going right and left, slighty up and down, in loop

[rough vessel]

oh yea 10k frames thats epic

[oak urchin]

The point is to follow the 120 frames duration per 'scene'

[oak urchin]

lmao

cadence 3 tho

So only 3333, i lied 😭

[rough vessel]

ill probably have to reinstall deforum it was working fine a few days ago now all videos become a blurry mess

[oak urchin]

ah, it may depends the motion/steps/noise yeah

and the anti-blur

Here : https://github.com/deforum-art/sd-webui-deforum/wiki/FAQ-&-Troubleshooting#my-animation-becomes-blurry-over-time-and-loses-details

[rough vessel]

yea i tried with different settings even the strenght scheduler and noise schedule but doesnt work something probably corrupted

[oak urchin]

Here, you can see the opposite. I increased too much the noise

If your video have fast moves and very slow moves, then you'll need to adjust these param during the video with a schedule yeah.
I'm almost done fighting the blur, i begin to understand

[rough vessel]

yea it was working fine before with the same settings but then suddenly something broke

[fallen girder]

Deforum Hybrid Motion (SDXL 1.0)
https://youtu.be/YziHGlBIyEE

[last holly]

ROBOTS doing surgury

[topaz snow]

billions of errors later

its still too flickery but i switched to ebsynth and spent most of my time trying to fix all the errors. now to make it smoother gonna go fix more errors

[rough vessel]

would u share the deforum settings .txt file u used for that?

[topaz snow]

i made that on ebsynth 😭

[rough vessel]

o i thought it was deforum

[topaz snow]

i made other stuff with deforum (posted somewhere up) but they felt too flickery so i switched to ebsynth

[rough vessel]

yea i have the same problem and controlnet too buggy with deforum prob will have to try with ebsynth then

[topaz snow]

yeah plus deforum was a lil bit too confusing for me ebsynth a lot more easier to work with but holy shit the amount of errors i ran into just to get it to run was a pain

it was mainly just me not knowing how to use ffmpeg lol but if you know how to use that and have everything installed correctly then you should be good!!

if you do end up trying ebsynth and need any help you can ping me! ill do my best to help

[rough vessel]

🫡

[nocturne crest]

I'm pretty familiar now with generating images with different models, played around with SDXL 1.0, etc. But what I really want to do turn my videos into animations like this: https://www.instagram.com/p/CvC3P9BteU6/

Is this stable diffusion? Do I just need Deforum? I'm at a loss where to go next. Any tips?

Instagram

[charred galleon]

No morphing, no diformers, rendered in dream studio, assembly timeline.

[void girder]

https://www.youtube.com/watch?v=nW8nlQEQ7cg

[lusty dagger]

i used mov2mov for this. is this ok?

[topaz snow]

thats looks pretty good

[cloud briar]

Why do the GIFs I generate with animatediff always have two sets together

[bold jewel]

I put a video to generate in the deforum, the video has 13 seconds, and it has been generating for two hours, a 920x500 video, is the delay time normal?

[high orchid]

3 beautiful anime girls in bikinis, next to f16, 4k resolution

[placid cargo]

https://media.discordapp.net/attachments/1101840261782450286/1139696124031139910/ezgif.com-gif-maker.gif

This is only little bro...

[topaz snow]

I think it depends on your gpu, what you are you running it on?

And also depends on the amount of frames and strength I think

[quick sonnet]

Any web or discord bot to generated single image to video with AI free ??

Espsecialy for NSFW image

[toxic glen]

I love it! Making horror animations is my favorite. Here's one I made using a friend's new checkpoint they just released

[toxic glen]

https://civitai.com/articles/554/dwh-1-creating-your-first-animation-with-deforum-in-1-click

I have three parts of my Deforum Tutorial series up so far, hopefully it can help you get started and understand the basics!

[topaz snow]

tried ebsynth with a longer clip

[vapid lagoon]

how did you make this?

[mossy ridge]

hey there, are you guys able to make animations using controlnet on the latest automatic1111 version?

[serene jay]

for sdxl no but there is an experimental controlnet canny one making the rounds (haven't tried yet)

[mossy ridge]

even with the 1.5, it's bugged

[serene jay]

1.5 controlnet runs perfectly with deforum

[mossy ridge]

maybe, but not with the batch

[serene jay]

please give me a little bit more context to try and understand the problem. also check that you are using the latest versions of controlnet and the models

[mossy ridge]

I was using everything new, fresh new install, trying to batch a folder with many pics using controlnet canny sd1.5, with it it doesnt work and gives an error, I just reinstalled the 1.4.1 a1111 and it works fine

[serene jay]

this is for deforum right? you have to give the controlnet a video or animation in mp4

not a series of pics

[mossy ridge]

never mentionned deforum, I don't use it

[rough vessel]

hes probably talking about the latest version of a1111 thats bugged with some old animation extensions

[mossy ridge]

it's a known issue apparently: https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/12001

[rough vessel]

just downgrade your a1111 version to fix it

[mossy ridge]

did it, it works

[serene jay]

oh I see, you are making an img2img batch?

[rough vessel]

downgrade to an old version like one from may

[mossy ridge]

(did: git clone -b 'v1.4.1' --depth 1 https://github.com/AUTOMATIC1111/stable-diffusion-webui.git)

for those interested

[jagged vapor]

https://youtu.be/mNsVV5wSL3A

Made a new animation using Deforum and SDXL

What do you guys think?

[topaz snow]

Looks sick

[faint rivet]

I redid an old one with SDXL refiner, with slight strength increase.
https://www.youtube.com/watch?v=A4k7feV-tAM

[void girder]

https://www.youtube.com/watch?v=vSPDJc7BGBA

[sage fiber]

I created the very first B-1 battle droid LoRA for this. Made an automation tool for ComfyUI that allows batch processing w/out deforum, ebesynth or even controlnet at all.

https://www.youtube.com/shorts/7Yc1Yr57TUo

[simple kiln]

Anyone tried to use animatediff with comfyui and 8gb GPU?

https://github.com/ArtVentureX/comfyui-animatediff

Found it

[oak urchin]

Lich Preview, still running (SOUND!)

Devilish Angels (SOUND!)

Cyborgish, just a test (SOUND!)

Made with SDXL, Deforum and Parseq

[mossy ridge]

Hello, very nice video, I like it very much.

[topaz snow]

Thanks dude

[lost dragon]

what web are u suing most for developing the designs

???

[bold jewel]

I made this animation using a tiktok video of a real person, but there is a doubt, does this person have the right to the video or something like that?

[chilly elm]

your majesty 👑 royal sloth 🦥💤
share some 🦥 love ❤💬 on the original post here: https://www.instagram.com/reel/Cv2UmwCuomH/

theairchitect

your majesty 👑 royal sloth 🦥💤

🤖🎥 A.I. Animation/Video by @theAIrchitect made with @stablediffusion SDXL @stability.ai
🔊🎶 music "Royalty" by Egzod, Maestro Chives, Neoni

#AIAnimation #AIArt #StableDiffusion #SDXL #Sloth #theAirchitect

Instagram

[dusk island]

You could try asking them if it's cool to post it, maybe? Perhaps going even further to link them in the post if it's cleared?

[fast basalt]

Emma watson hair transformation

[manic notch]

https://www.youtube.com/shorts/6p6frDXG4vU I've been playing around with a workflow in CompyUi trying to emulate the basics of hybrid video. This is a very far cry from the work of @faint rivet but it shows a step in the right direction. If nothing else, this shows just how good deforum is.

[timid wraith]

https://www.youtube.com/watch?v=MSzOgrMIpuY

[serene jay]

"Whispers of existence".
sdxl + comfyui + gen2 + midjourney + deforum + ae

[topaz snow]

damn thats alot of work, looks fricking sick!

[void girder]

https://www.youtube.com/watch?v=6qYXNAjdLrQ

[fallen girder]

Subway Chase | SDXL Deforum Hybrid motion
https://youtu.be/mUasqMQ8z1M

[oak urchin]

Extoplasm Lich

[oak urchin]

Fractales Zentangles City (SOUND!)

[dense crater]

https://www.youtube.com/watch?v=pzV_WLSpZm4

[toxic glen]

how many renders did it take before it didn't hive him boobas

horror animations are the best kind

[near badger]

https://civitai.com/models/105021
I made this lora preview prompt exactly as it is.
I got a normal video on the first try.
I noticed that the video generated by animatediff hardly has anything weird in it.

[toxic glen]

very cool!

[jagged vapor]

Made another song and animation using SDXL DreamShaper

https://www.youtube.com/watch?v=_mJw4IC7pNY

Appreciate any feedback! Thanks!

[fallen girder]

"Athena" | Deforum Hybrid Animation (SDXL 1.0) [4K]
https://youtu.be/4nGNndoXFEg

[jagged vapor]

Nice!

[faint rivet]

https://www.youtube.com/watch?v=upmzs96o2iI

[errant delta]

With my current methods it takes on average one and a half minutes per frame for a 608x1080 animation. Is this about average or is there a more efficient method? (Euler a: 25 steps on a 3060 with 4 control net nodes (tile, openpose, reference, temporalnet))

[void girder]

https://www.youtube.com/watch?v=tuZCoMxKJd0

[dense crater]

https://www.youtube.com/watch?v=hqsx7-c-qsY

[topaz snow]

yoo bruno animation?? 😭 thats so sick, i love it

[hot otter]

Wow, damnit I don't see why you can't make an anime with this, except the training costs

[hot otter]

Is your pfp from an anime or is it a gen

[topaz snow]

its from an older anime which i ran through the AI to make it look a little more modern-ish

[hot otter]

Do you remember the name

[topaz snow]

its from an anime called ergo proxy, the character's name is Re-L mayer

its pretty good, i recommend

[errant delta]

what do you guys use for deflickering?

[near badger]

https://github.com/s9roll7/animatediff-cli-prompt-travel

[high orchid]

create a dog eating an apple

[severe perch]

The animations are sick

[wanton panther]

Incredable! i'm sad I don't know enough about the pipline in code to use something like this yet cx

[toxic glen]

Great song, awesome video! I've actually found the -core metal genres (hard,metal,death) to be the most accepting and supportive groups of people when it comes to AI art (and none of it has anything to do with AI at all) A Deathcore band just released a vid2vid AI music video. I've even caught myself a music video commission .

i may have posted another version of this render here. but this is the original one i put on reddit and it got THE best question of all time

[toxic glen]

omg this is SSS tier

[wanton panther]

Here's the ComfyUI Animation Graph I've been working on. Still needs frame interpolation and a bell n' whistle but should be enough to get anyone off the ground. Posting this cause I haven't really found anything on Animation with ComfyUI. Shown Gifs are from the graph so it should work out of the box, once the custom nodes (Fizz, MTB, & Impact) are installed. If/When I work frame interpolation in I'll repost, regardless feel free to use etc as you wish. gl hf, happy generating~ ^^ <3

[dense crater]

Thanks! Yeah, they generally seem more accepting to most things, including AI art! Cool video and awesome response! 😄

[digital topaz]

portrait messi as in godfather movie, detailed, 4k, sharpened, colour graded, midtones, cinematic, cinematic view. ¯_(ツ)_/¯

[dense zenith]

My stupidest Deforum project. I spend a summer month locked in my room with overheating gpu and generated over 150 000 frames for this short animation.

Every source clip is upscaled to 5k 60fps and then color corrected, sharpened and finally downscaled to 4k (uhd).

https://youtu.be/Ge2vQ5a9OY0

[wanton panther]

So cool! whatever your techniques are def please keep them up that was a pleasure to watch cx

[hot otter]

@young canyon talk more about this impressive work

[serene jay]

really cool! did you check the fizzledorf nodes? rhttps://github.com/FizzleDorf/ComfyUI_FizzNodes

[young canyon]

I just created a workflow with deforum in blockeyAI. Its video to video with deforum and controlNet. Attack on titan LORA. Training costs not heavy, the real issue is its very low denoise. any higher and it flickering would become a problem.

[wanton panther]

Yee Fizz, i'm using Mtb's interpolater atm

[serene jay]

great! i understood that you weren't interpolating but we can see. are you diffusing every frame?

[wanton panther]

probably not

lemme get my current one

beep boop, this one has basic interpolation up if anyone wants it (i havent redon the comments and stuff just heads up)

[wanton panther]

ahh right diffusing = generating, each frame, so yee no, but i'll need to mess with it more to see what's works for me with the inter-frames from interpolation, vs frame rate, rate of the image transform and the "start_at_step" setting

[faint rivet]

https://www.youtube.com/watch?v=LCOkQFHBhCk

[somber lintel]

Hello everyone! the latest update to my Deforum-Helper app brings some new features. Your control options have expanded, now encompassing camera FOV, near and far. Plus, you can fine-tune the Deforum diffusion strength. I'd love to hear your thoughts on any additional settings you believe should be included in the editor.
I've also introduced a bar visualization in the 3D live mode for more effective control and monitoring as you work. Your feedback on these updates is highly valued!
Enjoy and thanks for using my app 🤘
https://deforum-helper.vercel.app

[void girder]

https://www.youtube.com/watch?v=ynR2MTJnTE8

[fallen girder]

"Virus Z" | Deforum Hybrid Animation (SDXL 1.0) [4K]
https://youtu.be/s9E25n_WVYs

[vital yacht]

2D animation created with Deforum + SDXL, no other tools, no interpolation.

https://youtu.be/edBZqJ0BR8g?si=O8GZgCataiiMFc7B

[wanton panther]

Progress with ComfiUI animation graph

[faint rivet]

“𝕯𝖊𝖛𝖎𝖔𝖚𝖘 𝐏𝖗𝖊𝖛𝖎𝖔𝖚𝖘 𝕱𝖑𝖔𝖜 𝕱𝖎𝖌𝖍𝖙”
https://www.youtube.com/watch?v=WCkcDCHWQVY

[faint rivet]

very deforum-like!

[meager scarab]

Hi Everyone, I made a CV video for myself with AI art. Enjoy watching..
https://www.youtube.com/watch?v=twxzOu0YDM8

[stray tree]

"Mechanical art🖌️ "
https://www.youtube.com/watch?v=oS50BlwfZEg

[languid badge]

Could someone point me to some information on how the Deforum batch mode works?

[spare siren]

hola

moving guitar change color

[marble shard]

"It was beautiful". Made with SD
https://youtu.be/qWbtYGFIMc8

[restive ocean]

https://media4.giphy.com/media/ng8bihkFt0dDAc8aOb/giphy.gif

[wanton panther]

Just a boring scroll test I ran overnight, anyone know how to walk between prompts in ComfiUI?

[void girder]

https://www.youtube.com/watch?v=XDqg_yDSo20