▶｜stable-video-diffusion

[odd turret]

made with pika lab. But any image can be generated with SVD.

[rancid verge]

Guys is it possible to fine tune stable video diffusion

I was thinking about fine tuning it on steamboat Willy

As it recently became public domain

[tribal jackal]

Has anyone run into this code "fetching new files" indefinitely? Used to work fine but stopped working recently.

    "stabilityai/stable-video-diffusion-img2vid-xt", torch_dtype=torch.float16, variant="fp16"
)```

[steep creek]

A boy play football

[onyx silo]

Are you working in colab, or maybe another notebook , there's an multi-threading issue with authentication with huggingface, the recommendation is to update huggingface hub.

# run before executing any code
!pip install -U huggingface_hub

[latent moss]

https://www.midjourney.com

a painting of a beyond-dimensional fantasy world,reality infused with dreams,in a screenshot from the science fiction anime file by makoto shinkai

[median loom]

if you can´t create images locally, you can use #1100170312106127410 , or some online websites like https://creator.nightcafe.studio or https://tensor.art/images/680095568950906397?post_id=680095568946712095&source_id=nzm-pljkk0W0pfQvbHb29Rcu

[odd turret]

still dont understand how parameters work. I watch some tuto where people put 20 init images to get what they want. Not sure it is what I want to do...

it seems that things work with a library of videos. Therefore, when the init image is easy to understand for ai. It will give good result. Like a car driving, a boat sailing, face of someone. But if it is too weird, it just does nothing...

https://tenor.com/view/robot-bully-gif-20161386

stupid ai...

[ivory ferry]

A boy play game

[glacial obsidian]

Are there any controlnet type models for stable video already?

[drifting vortex]

There is one, yes.
https://github.com/CiaraStrawberry/svd-temporal-controlnet

[open heron]

https://github.com/ProjectNUWA/DragNUWA

[tribal jackal]

How do you avoid videos that look like this?

My settings:
num_inference_steps = 10
decode_chunk_size = 12
motion_bucket_id = 50
noise_aug_strength = 0.02
min_guidance_scale = 1
max_guidance_scale = 3
resolution = 512 x 512

[bright kraken]

What is your cfg setting?

[keen shuttle]

https://youtube.com/shorts/lnEUBXpX4LA?feature=share

[viscid mortar]

can you guys tell me how i can use Stable Diffusion Video ? Do i need to install it on my mac ? is it hard to install ?

[keen shuttle]

I think you need a PC with a nvidia gpu with 8gb of vram minimum specs

[viscid mortar]

good to know. I just have a mac laptop. Is there any installes on servers online you can use ?

[keen shuttle]

@viscid mortar not for SDV yet,looks like you can get on a waiting list. But theres "Pika AI" and "Gen2 by Runway" that offer online server for Video outputs

https://stability.ai/stable-video

[honest dagger]

will able to use stable diffusion videos it my RTX 3050 (75W) Laptop

4 gb ram

soryy 4gb vram

[odd turret]

https://civitai.com/models/261438?modelVersionId=294910

[atomic sphinx]

https://www.youtube.com/watch?v=GCYUnyMe4OI

[atomic sphinx]

https://www.youtube.com/watch?v=8u-2eX5NMmg

[hybrid rose]

When you make an AI image into an AI generated video.

I tried Euler then tried DPM2

I like the results with Euler

Euler with DDIM_uniform scheduler

[hybrid rose]

Original image I created with SDXL and Image to video with SDV

[hybrid rose]

This is wacky. It does anime very well.

[glad marsh]

https://huggingface.co/TencentARC/MotionCtrl/blob/main/motionctrl_svd.ckpt

[rose pivot]

trying a music video with stable diffusion 0.7.1

https://youtu.be/GcUREkigAkA

[bright kraken]

Have you tested it with SVD yet?

[rancid basin]

We built SVD in to the Beta version of our app (private non-commercial), but if anyone wants to check it out it's great for doing Storyboard Previz, here's a sneak peak: https://youtu.be/hukP5nyvfP0

(our app SAGA also helps with Storyboards using SDXL, and a variety of screenplay writing stuff like loglines/plots/characters/acts/beats/scripts)

[glad marsh]

👍

[glad marsh]

camera motion is pretty good in some case

[final spindle]

Hi guys, I made this cool animation..
First made 50000 frames, then interpolated it into 200k.
However, I'm not 100% happy with the smoothness and the few artifacts around the corners.
--> I was wondering how to get a smoother video?
--> I'm looking for recommendations on the best practice workflow and settings.
Workflows with ComfyUi or postprocess Flowframes/Davinci/other software settings are all welcome!
Many Thanks!! --
https://youtu.be/Qr7PJA3X65U?si=5M5_8sWHt064rkS8

[atomic sphinx]

https://youtube.com/shorts/4DHVipv9jMY?feature=share

[icy pendant]

1850's newspaper printing press, typeset letters in slot, --ar 16:9

[atomic sphinx]

https://youtube.com/shorts/4ZfUJ9JIAG8?feature=share

[silent hinge]

I am new to Stable Diffusion. I am interested in working on it. I also tried SVD and SVDXL on Collab. But it isnt working due to some errors. I am trying to debug but not able to get over it. Can someone share how you were able to achieve these results. Any relevant code that I can implement?

[iron topaz]

Can I run SVD on my 3070 8g laptop?

i712700h, 32gb ram, laptop rtx3070 8g, 2t storage

[unkempt mica]

Test it to see but I dont see why not. I have run it on a 2060 super but its super slow.

[queen jacinth]

I quite like the result here

[uncut sigil]

https://youtube.com/shorts/xFZ5WfGRidE?si=PYWbjL4MoN-Busmo

[red fossil]

is there anything written up on the "deflickering decoder" that SVD uses?

[chrome bolt]

I want to use img2img to generate an animated image from the original image. Is there any plug-in that can do this?

[tawny star]

hey everyone, lots of SVD videos 👇
https://www.youtube.com/watch?v=8eJVbiXEpNU

[tawny star]

if you're using stable diffusion, there is an extension to do exactly that. search 'deforum' in the extensions.. this is the repo
https://github.com/deforum-art/sd-webui-deforum

[limpid horizon]

Hi is 2seconds max for svd now?

[tawny star]

its more about 'max' frames which can still be increased but the issue is theyre trained on 16 frames and set to 8fps.. you can interpolate and 'stretch' it to get a bit more to only to some extent

[chrome bolt]

Ok I will try this plugin thank you very much

[atomic sphinx]

https://youtu.be/q2PoQAW_VP4

[chrome bolt]

Sorry, I think I expressed it wrong. The effect I want to achieve is to generate a gif from a picture. But the current technology seems to only support generation through prompt.

[tawny star]

there are many that do that too! https://github.com/guoyww/animatediff/ you can install it as a Stable Diffusion extension, or https://github.com/Vchitect/SEINE

👈 check out my discord 🙂

[gentle trellis]

I realized that the last frame of SVD is super saturated. Why does it happen?

[junior edge]

Try lowering CFG a bit

[winged garnet]

gitl

[elder hatch]

when using SVD in comfy UI how many steps do you guys typically use? I started with the default in the workflow I had which was 12 steps and euler and after bumping it up to 18 steps I noticed a nice improvement

also I've been able to get decent video bumping it up to 40 frames

sometimes it falls apart though

is this SVD usually it does a little more than just zooming in and out

[nocturne lake]

Sorry, it's not good enough, this needs to get faster. The speed of quality images taken per second should be increased. We need an even faster sdxl turbo.

[elder hatch]

no its fine usually video with SVD has the character do something. Sometimes it does just the zooming effect but if you run it with some different settings or re run it you should get some more interesting movement

[hallow cedar]

When is the model for SVD 1.1 available?

[slim nebula]

Pandas sway their umbrellas and jump happily

[mild mason]

mind blown...

[mild mason]

Are there any available that are built around sdxl as a base model? I see this one mentioned within the animated_diff repo you mentioned, are there others?

https://github.com/guoyww/AnimateDiff/tree/sdxl

Maybe someone is working on integrating this into fooocus...

which would be awesome

[bleak apex]

oh look, new version https://huggingface.co/stabilityai/stable-video-diffusion-img2vid-xt-1-1

[dusk dagger]

https://tenor.com/view/andy-bernard-the-office-dance-celebrate-woo-gif-16554183

[coral heart]

Wow and the size is only 4.78 GB, should take less vram, thx

[hoary heart]

Does anyone know how to make it faster?

It takes 45 sec for me

like can it be optimized with lcm

[open heron]

Smaller image and fewer steps

[lucid wind]

Hello friends, I hope you are well, I wish you a happy weekend, does anyone know if there is a Google colab to use animated diff?I don't have a good computer at the moment but I would like to try it Thank you all 🙏🏽🙏🏽

[mild mason]

it took me two days... but I finally got SEINE one to work on windows

!!

the sample image works but not any old image from your files... it needed to be converted to RBG

[jovial sparrow]

Did you ever find anything better than Flowframes for interpolation?

[open heron]

SVD 1.1 is simply better than SVD

[glacial obsidian]

has anyone tried finetuning it for anime

[open heron]

I don't think it can be finetuned

[severe moon]

new version seems faster, nice

[severe moon]

I wonder if they trained it on more animation stuff, seems to do better with art beyond just trying to pan over it like the older models

[severe moon]

what are you lookin at?

[lean dragon]

https://tenor.com/hyIcPFRkfIj.gif

[brave shard]

Others/dream prompt:love song style:Photographic aspect:9:19 format:Image negative_prompt:

what are you lookin at?

[ebon flax]

generate image

[unkempt mica]

ill do my best

[left zenith]

Has anyone found a way to load the new SVD 1.1 checkpoint with Diffusers?

[fresh tundra]

Can I use Stable Diffusion Video in Invoke AI

[wispy garnet]

Stable Video Diffusion 1.1 first trial
https://youtu.be/xgeRSi1PdEo
(performed on Google Colab A100)This video includes some failed cases.

The overall impression is more stable than the first release. Mr. Mkshing pointed out that things like the logo are completely stable. It's as if they are treated as separate layers. Also, SVD1.0 was not very good at generating anime like digital illustrations, with only panning, But SVD1.1 sometimes generates them well after 2-3 times. We recommend trying the same input several times...!

Ask X(Twitter)@AICUai for more details!

[atomic sphinx]

https://www.youtube.com/watch?v=AuxJfhM90XY

[winged drum]

holy, which UI are u using?

[coral heart]

The usual, A1111 for image, ComfyUI for video (dafault image to video workflow)

[torn osprey]

looking to create pretty specific animation videos (~5-10 sec) like patient animation or car crash animation with type of car on a certain crosswalk, how should i approach this with the latest available tech?

edit: automated if possible (or a code workflow I can use with minimal manual intervention)

[lapis field]

https://www.youtube.com/watch?v=dBypbWk7BXU

Music in Suno

Images and Video in SVD (images produced first in various checkpoints, then passed into SVD and interpolation)

Lyrics mine.

[south wave]

Hey, guys. I've just developed a new system. I'd like to read some thoughts on it, if possible ♥ https://www.youtube.com/watch?v=zptPQbTScto

[coral heart]

🤣

[bleak apex]

i just say i did it on purpose

[dusk dagger]

is it just me or do the example results for the 1.1 model seem unattainable?
https://huggingface.co/vdo/stable-video-diffusion-img2vid-xt-1-1/blob/main/svd11.webp
seems like more than 25 frames of effective context

wonder if they interpolated them with some other tool, or even SVD itself in some non-publicized way?

[bleak apex]

dunno, interpolation is not magic sauce, probably more along the lines of seed hunting/lotto

the cat video i posted up there is interpolated right in ComfyUI 4x

using really good inputs would probably help me, im kinda just throwing trash at it

[dusk dagger]

what exactly is interpolating it there? looks pretty good

I've used FILM for interpolating my AnimateDiff outputs but I'm not sure I like the effect it makes a lot of the time

I would hope Stability isn't using something like FILM on their example outputs

just feels like more than 25 frames to me which as far as I know, is the max context length SVD can do?

[bleak apex]

that was rife 49 i think

[tawny star]

cool!! I have tried that one and couldnt get it got work. I did get Lavie work, their other repo. And I'm not sure which of those base models they use, I use the AnimDiff for automatic 1111's SD. Kagi below your messed posted their high def model for AnimDiff

[mint forum]

hey guys - what are the minimum requirements for my PC to run SD on my Mac

[pure wedge]

How to use it?Did the svd1.1 is similar with stable diffusion, Or they are in same channel

[queen frigate]

是不是永不了了

是不是用不了了 找不到机器人 了

[sturdy shuttle]

为何突然没有权限了

[nocturne lake]

https://youtu.be/9Z406wtJ1v8

[vagrant dagger]

能用AI生成一个用虾摆盘的图片吗 摆成个龙的形

[spice sapphire]

当然可以，只要你提供详细的prompt

[atomic sphinx]

https://youtube.com/shorts/Qxdr4g4ZULE?feature=share

[shy ether]

Hey Guys, I tried downloading controlnet but for some reason it doesnt show up on my deforum, is there any reason as to why?

[distant kernel]

#🤝｜tech-support

[paper shoal]

Does anyone know if Stable Video will have a self-hosted version, similar to Stable Studio?

[tawny star]

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

[molten latch]

where do i go to start generating images with prompts .. a lot has changed here

[soft stratus]

Right now you don't. #1047610792226340935

[late umbra]

ai cars

[ocean drum]

Virtual shooting based on SVD ComfyUI custom workflow on my music (JEFF3000)

[lone mortar]

generate image: COW WEEK #23- 1st Place & Staff Pick- "Stable Pets"

[brittle solstice]

Howwwww??? Local install?

[severe moon]

yes - this was made with just a comfyui workflow running locally using the latest svd_xt 1.1 model on a Midjourney image.

[brittle solstice]

Love it

[heady gull]

make magic

[noble pumice]

SVD is magic https://youtu.be/NU510gUtD4M

[steel stream]

this roman legionary and the cloak floating while he is moving! dope!

[noble pumice]

Thank you! Yes the SVD model is the only thing I've seen to replicate natural walking movements 💫

[steel stream]

were these generated from scratch or did you have images as inputs?

[noble pumice]

99% of the reel above is Midjourney Image to Video with SVD. Then everything upscaled with Topaz Video AI to 4K 60 FPS.

[steel stream]

👏

[noble pumice]

By far the most superior model for natural human movement

[steel stream]

agree!

[primal jolt]

How To use this

[spring portal]

if only they'd open up their dans BETA. Been waiting forever and resubmitted again. I can't deal with Gen2 no more SAVE ME SVD

[fossil merlin]

By far the most superior model for natural human movement

[ashen tusk]

Perhaps the second most superior model now? This is just unhinged...
https://openai.com/sora

[silent hinge]

stable video is like a cave drawing to sora's academy award winning film, not even in the same universe

[unkempt mica]

Then go run that model.... wait, you can't....

[ocean needle]

Sora at least shows us what's possible now. It's crazy good.

[silent hinge]

not yet, but i guess i can go squint and watch a jpeg flicker and stutter and call it stable video while i wait

[unkempt mica]

I think you missed my point. You can never run that model its private. You will never be able to do whatever you want with it unless you have an enterprise account. Even when they allow you to use the one on their servers it will be heavily censored.

[ocean needle]

Nothing to worry, open source model will catch up soon

stable video diffusion was released 7 months ago, I believe there are more unreleased research we haven't seen.

[unkempt mica]

I think its inevitable that it will catch up. idk about soon. We still have not been able to get an LLM thats as good as GPT 4

[silent hinge]

not really, im not lookng to make violent stuffs or porn. people say this about stable image too, but it pale in comparison to mj and dalle 3 and suits my needs

[ashen tusk]

While I agree with the sentiment - I must disagree with regards to Dall-E3, which to me is just garbage. Surely, the architecture and the model itself is probably awesome, but it is unusable and without access to the underlying settings the output is just sub-par. I'd argue, at least in my use case, SD, due to its active community, is catching up to MJ as well in terms of quality of output, and by far beats MJ in terms of speed while testing.

[ashen tusk]

What I would like to think is that we will see specialized models implementing the same or - within some boundaries - similar implementation of foresight to secure consistency over time. Obviously there are limitations in the availability of compute and VRAM for the vast majority of amateurs and enthusiasts (for example when it comes to render a full 60 second clip all at ones), but that, then, could also be a driver for invention and development i guess. Anyhow, Sora surely is something highly interesting, it feels, to me, as a groundbreaking leap beyond SotA, rather than just incremental. I do acknowledge that this idea might reflect my lack of understanding of the underlying architecture though.

[ocean drum]

I use a lot SVD but Sora is really a game changer : vid2vid, img2img, text2img, camera handling with prompt, epic consistency, overall quality and 60 seconds generation... 60 sec... it feels like everything i learned so far in video generation will be soon pointless... I know that all SD preprocessors will give us more control over generation but OpenAI new tool makes me dizzy Have a nice day!

[silent hinge]

we must be using different dalle3. its the best imo, only some photoreal stuff is throttled but even there it can beat mj a lot of times. try any prompt in dalle3 on bing image creator vs any stable and dalle wins every time imo, save for here of course where people are stable fanboys and get all mad when you say otherwise

and you know sora does images too

[unkempt mica]

@silent hinge please don't misunderstand anything I have said. The OpenAI products are better than SD, better than every other models out there. Sora is better than SVD.

When mentoned censored you went straight to thinking I was talking about violence or porn. You definitely cannot make things like that (personally I like the horror genre and a lot of that is blocked) but once I asked DALL-E to make me a meme of a doctor but was told it was against the guidelines. Some people get stopped from making an image if they ask it to be a girl at the beach.

Then there is the usability and price. Maybe your workflow is less than mine but I make bots that generate images for my friends and family. I make hundreds of images per day dozens of SVD movies. Doing that in DALL-E would be an insane cost and it would only allow me to generate so many.

With the open source models I can do whatever I want whenever I want and I'm only slowed by my own hardware limitations.

To sum up my opinion on this. I love that they are making this new video model. And I'm sure I will spend some money on the API when it's available. I like to see what currently possible and hope the open source models come close soon.

[elder hatch]

Emad on reddit said stability has something they're working on in the ballpark of what openai is doing woth Sora but implied they need more compute power training it

Don't have the link to the comment

[wise forge]

Hello everyone, please I have a question, still a newbie. With the stable diffusion image to image, please could anyone explain why the generate image does not match the body size of the in put.

[dreamy coyote]

Is someone using Automatic1111, and what are you using for generating video?

[viscid beacon]

Hello Michael! You need an extension called "ControlNet" for it

https://www.nextdiffusion.ai/tutorials/how-to-install-controlnet-extension-in-stable-diffusion-a1111

@wise forge

[wise forge]

Hello, thanks very much, I will try to follow the article

[viscid beacon]

👍🌟

[steel herald]

im using image to video. how do i prevent weird hand animation (cfg_scale 1.8, motion_bucket_id 128) . im using the stability ai api (https://platform.stability.ai/docs/api-reference#tag/v2alphageneration)

[distant heron]

#📝｜prompting-help

[shrewd shadow]

Hi. Good luck guys. Where can I find prompts suitable for the furniture or can you help me?

[alpine shard]

生成视频

[coral heart]

Could they make it even more censored? I'm starting to hate that "dog" picture with a passion 😄

[pulsar finch]

https://github.com/TencentARC/MotionCtrl anyone had luck getting in the queue for this?

[atomic wing]

Could somebody please link a tutorial on how to install on Mac

[atomic wing]

So when sora comes out you will buy its api to make a bot that uses it?

[unkempt mica]

I guess it depends on the price.

[atomic wing]

Day 2: could somebody please link a tutorial on how to install SVD in Mac?

[south kettle]

Hello! I would like to create text2vid with a speaker, similar to this: https://www.deepreel.com/

Do you think they use SVD? Do you have any tips on how I can get started?

[peak rampart]

a boy is running

[worthy hornet]

By far the most superior model for natural human movement

[steel stream]

guys, are you using the API video generation or is there some other tool within SD for videos?

[bleak apex]

anyone get AnimateLCM SVD working, i was failing last night just getting black output trying to figure out how to use it, i did get the plain AnimateLCM working though

i think i answered my own question

[silent hinge]

Hey guys I was wondering if someone here has any experience with creating a custom spatiotemporal controlnet and finetuning the SVD

[tawny star]

https://www.youtube.com/watch?v=jYdqz5443dw

[rain owl]

very often the video that is generated is very blurry or deepfried. Does anyone have tips / settings to share? the two workflows i've attached are the ones i've been using, mostly unsuccessfully

[fast tundra]

great job

[celest elbow]

Hello! I was wondering what anime model was used for this video! Been looking for some good anime models and this looks really good so I want to try it out, and if anyone got el recommendations I’d love to take a look!

https://youtube.com/shorts/gbt-LQKv0hg?si=Ud76z7smgX1bkJir

[frosty raven]

So what SVD settings are you guys using to get useful output out of SVD? I've tried 10-15 generations with the settings I got in the screenshot, which I got off a tutorial on civitai. I only get useless gibberish with characters just flailing all of the place. I haven't gotten a single output that I'd want to show anyone else. Ideas? Thanks.

[fallen wren]

if you use the Video parameter option in StableSwarmUI, and import to Comfy Workflow Editor tab, you'll see the settings Swarm uses for SVD - has custom sampler/sigma/etc. (SVD responds well to much larger sigma value)

[frosty raven]

ok thanks I'll have a look

[silk verge]

Hello. I've been looking for an answer to a question for a long time. Video that transitions between 2 specific images. Switching between 2 specific images previously produced in Photoshop without changing the style of the images. Creating a video that transitions from photo A to photo B with artificial intelligence. From Stills to Motion. Can you help me?

[silver mortar]

hh

[hearty creek]

There's a decent learning curve to it but I think the comfyui animation ecosystem is the cutting edge for that sort of thing ATM. For what you're trying to do i'd check out steerable motion. https://github.com/banodoco/Steerable-Motion

[radiant nymph]

How to use this channel

[west tulip]

okay i know the video is horrible but i didnt know svd could handle things going into water. anyone have a better example of this actually working? 40steps. I can share my workflow later

[bleak apex]

closest i have

[west tulip]

here's the workflow incase anyone's interested. I'd love to see this with the head animated properly

[celest elbow]

Anyone what model/vae etc was used for these videos?

would greatly appreciate it if anyone can help me find out!

[west tulip]

how much vram do you use up with SVD? I max out my 12gb easy...

[tribal hatch]

Maxing out at like 19gb

[silent hinge]

discord cant display webp?

[silent hinge]

15gb here

oooo

how long does it take to generate a video at 25 frames with SVD-xt 1.1 for you guys

its 45 seconds on a 4090 for me

https://cdn.discordapp.com/attachments/1138803599640641547/1211175386613026836/svd-xt1.1_00018.mp4?ex=65ed3dfa&is=65dac8fa&hm=0c5ebb89e81bcf7531a7354e21dbc5dad7aefdfba1f4c8827ce2f8a33a8642a8&

[silent hinge]

lets get trippy mane

[winged drum]

whats ur CFG value on this one

[winged drum]

and here

[silent hinge]

idk

2?

2.5 or so

[winged drum]

yeah so standard right

[silent hinge]

yea

[winged drum]

i need to render 4-5 clips till i get the shot i want - searching for some controling nodes but cant find anything

so i haver tu burn ,my 3090 ti all the time lmfao

[winged drum]

btw 45 secs is fast

iam at 1min 20 secs for 25fps

and if i tweak stuff iam around 2:30

but iam about to buy a 4090 next week so yeah

[silent hinge]

what are you running now

[winged drum]

3090 Ti

[winged drum]

or iam running cloud distributed gpu´s

iam having a call tomorrow with some cloud services

[winged drum]

https://www.reddit.com/r/comfyui/comments/18zktfh/network_distributed_rendering_on_4_gpus/

look at this

lmfao

[silent hinge]

why not just keep using that until september when 50xx series comes out

its only like half a year

not worth getting a 4090 now imho when it loses like 30% value in 6 months

[winged drum]

its basically stupid to buy new ones at the end its all about the vram

also thinking to buy two 3090´s on top of my 3090 ti

will decide next week

u can buy a 3090 for 700 bucks now

used

who gives a shit

[late spear]

How do you make stable diff video stop producing garbage like this? Not one workflow I've tried with comfy UI works.

[sonic viper]

https://www.instagram.com/reel/C3zczTfJJbe/?igsh=MzRlODBiNWFlZA==

manish_minglani

The earth has its music for those who will listen;
Its bright variations forever abound.
With all of the wonders that God has bequeathed us,
There’s nothing that thrills like the magic of sound.

~ Reginald Holmes, as part of his poem, The Magic of Sound.

🎥 AI Inspiration for video comes from the beautiful piece done by @urbantheory_

#nature ...

Instagram

[quasi lantern]

35 minutes on m3

[silent hinge]

m1?

[quasi lantern]

m3 max

sry

[glacial obsidian]

Is there any official training code published

I tried using SVD Xtend but i cant get it to work

[fleet gorge]

https://mp.weixin.qq.com/s/U7QEsVKMSkplfUBw4hrn4Q

[pale osprey]

Hello, If someone can address my question regarding animatediff, I've encountered a limitation with animatediff model v3 where I couldn't exceed a batch size of 32. Previously, I was capable of creating them with frames of 120 or even higher but not anymore?

[pale osprey]

for example of my my video using mm_sd_v15.ckpt isn't behaving like it used to. I'm starting to think I might've missed something important-- I should have written it down.

I just did a quick search, so this might not be the best example, but it gives you an idea of how long it lasted.

[clever comet]

Can someone point me in the right direction -- 1. Is there a SVD 1.1 Extended version, like is in SVD? I usually just manipulate the fps etc and get a general output of 5 seconds. 2. I have seen the ability via YT to get camera control, is that available to us yet? If so, can someone point me to that link. I realize I am asking a lot, if anyone could help, that would be just grand! Thank you!!!

[pale osprey]

As what I know that camera control goes with Deforum with a camera movements.

https://civitai.com/articles/248/deforum-camera-animator

[clever comet]

Thank you very much!!!

[elder hatch]

dang when you're running a whole batch of images and one of them actually gave you a realistic blink

got about 80+ images im running, then plug them into a video editor to a song i made then attach the lip sync

[glad marsh]

https://github.com/TIGER-AI-Lab/ConsistI2V

https://humanaigc.github.io/emote-portrait-alive/

[trail atlas]

Who guess this anime?

[glad marsh]

sora arxiv: https://arxiv.org/pdf/2402.17177v1.pdf

[celest elbow]

does anyone have any model reccomendation for really good anime renders? and does anyone know why my renders have been so slow all of a sudden? super confused

[celest elbow]

[inland cedar]

use mf ghost

[shell vapor]

Today's tests show video output is taking much longer than usual, and it seems the computational power for dynamic parts might have diminished compared to before?

[frigid harness]

I made a video a couple months back on YT with regards to integrating genai into VFX workflows. At the time there wasn’t much you could do with temporal stability for locked down camera motions. Currently I’ve been working on training motion lora’s from AnimateDiff MotionDirector (DiffDirector) with CG inputs. I haven’t been able to get the motion of all the elements such as the ground working yet but it’s only been a couple days and there is a lot of improvement on tweaking the lora training and the cg source that goes into it.

Here's with no temporal stability just rendering out a frame sequence form diffusion. https://youtu.be/88uv9-PVJbY?si=2yrPRLjhTq3rwawN

More about the workflow with CG elements I cover in this longer video for still images only. https://youtu.be/lFE8yI4i0Yw?si=BaWgNXIb6iftGDhm

I'm keeping with this robot asset because it's something that no diffusion model has ever been trained on, and the constraint is the camera motion is locked down can can't be changed. Every frame (120) has to be generated so no doing 1/4 frames and interpolating between them.

If anyone else has tried training loras on cg motion let me know if you've got any suggestions to it. I could probably get better results if I reversed the frame sequence so the camera/robot are moving away from each other, but would like to find more robust methods as camera motions might never be ideal in this sense.

[brave nova]

what settings did you use for this to get the video to not be blurry

[ruby geyser]

I made this anime video inspired by 'Attack on Titan' on Stable Video. The script, sounds, dialogues and music are also made with various AI models. You can watch it in better quality here: https://aiflixhub.com/movie/shadows-of-freedom-1 #🍥｜anime #▶｜stable-video-diffusion #✨｜sdxl

[charred cypress]

https://youtu.be/Ep0SlzTlUg4?si=cI3Pp3IZ_OktRTBv

[cerulean fox]

Been playing with processes a lot and one I've come up with is taking and making audio with Bark, using that to make an low quality animated 3D video using the audio I made. Take that into ComfyUI and use the video with AnimateDiff in a pretty massive workflow to augment the video with AI. Then take the resulting video into automatic1111 and use Wave2lip to fix the missing mouth movements. Not Stable Video, but features I'd like in Stable video and this uses Stable diffusion so 😜

[finite bluff]

1gril

[night sparrow]

？？

[silent hinge]

||A iridescent, futuristic female android wears a ivory costume, patrick nagel, perry rhodan, john trumbull in the style of anaglyph filter, robotic expressionism, rainbowcore||

[wise star]

sharing some of my work using Stable Diffusion

https://x.com/cantavfx/status/1762904378082779200?s=20

canta vfx (@cantavfx) on X

spiderversing spiderman.. 😂 #aicreators #ai #aiart #aiartist #aivideo #aianimation #aifilm #aiartcommunity #midjourney #midjourneycommunity #EveryDream2trainer #stableDiffusion #ControlNet #SpiderMan #spiderverse

Twitter · Feb 28, 2024, 6:15 PM

i also have some breackdown
https://x.com/cantavfx/status/1762244422543290382?s=20

canta vfx (@cantavfx) on X

Zombifying #Aquaman with breackdown! @cantavfx #aicreators #ai #aiart #aiartist #aivideo #aiartcommunity #stableDiffusion #midjourney #aifilm #JasonMomoa #zombies

Twitter · Feb 26, 2024, 10:33 PM

[dapper spindle]

does anyone know how to use inpainting on video footage? is this even possible?

[cerulean fox]

Madeb with Stable Video Diffusion

[cerulean fox]

Automatic1111 has replacer. It's inpainting for video.

[strong bone]

帮我画，a peeled orange with one segment replaced by a clove of garlic, The garlic clove is nestled among the orange segments, creating a striking visual contrast.

[flat whale]

Here is the image you requested.

[rugged sequoia]

Help me draw the road map to Mars

[analog flare]

A guy robbing a store

[main mist]

on the fly 🙂

[glass copper]

paint a scene of cultural revolution that happened in china 1967

[ocean atlas]

Is Stable Video Diffusion then its own distinct model?

[fathom nimbus]

Here is the image you requested.

[urban shore]

We cant create image now?

[sudden finch]

?

[cerulean fox]

Yeah, Forge fork of webui has one of the better SVD extensions right now. ComfyUI has some decent workflows for it as well, but they can be janky at the moment. Forge is the best for it currently for local users.

[fresh tundra]

Is it possible to enter a prompt for Stable Diffusion Video or does it only work with images? I tried it but it seems to be very difficult how it works

I used Pinokio

[ocean atlas]

what's Pinokio?

[fresh tundra]

https://pinokio.computer/

[ocean atlas]

So I was wanting to get a look at the code, and even the underlying model, to see how Stable Video Diffusion works.
I don't really care about web gui's, since I prefer to run python code directly.
But so why is SVD called an extension? Is it something built on top of regular SD?

[cerulean fox]

Stable Video Diffusion is its own model for video generation from Stable Diffusion. The only way to run it locally is to either use an extension or custom nodes in an existing UI or setup your own system. It doesn't work the same way Stable diffusion does and the model is a bit Different.

I used SVD in my pipeline, so I'll post it here. It's another daily challenge I did for a group.

[ocean atlas]

Yes, thanks - I already watched it - great stuff
What is Stable Audio? What does it take as input, and what does it produce as output?

[cerulean fox]

Stable Audio I haven't played with as much. It's evolved a lot since I checked it out. It's an audio generative system, think the HarmonAI guys worked on it. Good team, I just got hooked on Spectrograms and went with other tools. I knew Spectrograms would eventually lead to voice and they did lol. But Riffusion is a Stable Diffusion model, it's trained on Spectrograms. You can use it in some repos for Stable Diffusion, the code is very outdated. Audacity is better for Riffusion local or their website, they can do vocals now as well on their site. Suno AI is another good one. After Chirpy they started being able to do music so they have their own music Gen with vocals now. They made Bark. Stable Audio can do some good sounds and sound effects, but I don't think they have vocals yet, but the Audio and sound effects are pretty good. I'm local only though because I want to experiment with everything all at once and I can't afford all these sites lol.

[ocean atlas]

So Stable Audio can do voice? (just not musical vocals perhaps?)

It produces voice from a text prompt?

Can you control the emotional tone/pitch of the generated voice?

[cerulean fox]

Sorry, I meant I don't think it can do voice. Suno AI can, but not Stable Audio, not to my knowledge at least.

[copper fulcrum]

Hello my name is Camilo from Colombia, actually im in a personal challenge, my team asked to created a anim with this kind of character

[empty flower]

bro who added half life 1 texture quality to one of the rocket engines

[lime canopy]

will there be another roll out for this? id like to test it out

[main mist]

Did you let the default motion setting at 127 ? I am not an expert, but the few I have played around with SVD gave better results <50.

Also, the default sampler is Euler which is faster but less accurate, tends more to deviate from original image.

[cerulean fox]

It's kind of neat messing with the motion settings. You can get Different animations tweaking it.

[worthy raptor]

Hello, I am looking a uiless solution for image to video generation, and Ihave a few questions about animatediff:

• Is there documentation somewhere, explaining configuration fields, folder structure, etc?
• Text to video generations via the provided examples with a resolution less than 256x256 (128, 64, 32, ...) produce garbage. Is this an inherent requirement or am I missing some config?
• has anyone tried with the image2video functionality fork (https://github.com/talesofai/AnimateDiff ) ? If so, does model loading work in the provided example for you?
• Do you recommend other solutions for image to video gen?
  Thanks!

[main mist]

Do you mean AnimateDiff ? Sorry, not sure what you are referring too

[cerulean fox]

SVD, there's a wide range of different motions you can access by adjusting the motion settings. I was agreeing with you about not leaving it default.

[main mist]

Yes 127 seems just too high to get a consistent suit of images. Do you tweak other settings too ? I haven't yet experimented with these, esxcept changing samplers to see what differece it makes (beside processing time, I don't yet understand the impact of selecting other samplers than Euler).

[cerulean fox]

Sampling method can have a huge impact on gens. Definitely expirement with a fixed seed so you can see the differences. Numver of Steps can change it drastically as well.

[ocean atlas]

I am also trying this right now, and am getting lots of memory overflow errors ("OutOfMemory")
I have 8 GB of vram on my video card -- how much VRAM do you have? Are you getting similar errors?

[ocean atlas]

Is it possible to run SVD on the cloud?

[main mist]

Thanks, you are right, best thing is to experiment around.

[soft stratus]

Here is your image:

[normal veldt]

Can anyone help me out with some stiching? I'm trying to see if there's an easy way to maintain a steady brightness in the background as I "lenghten" the video?

[twilit nova]

Hey guys! I am new to this community. Can someone tell me if these videos shared in this channel were generated by AI? If so, which model?

[abstract zenith]

/gif dog

[long halo]

Thus kind of Ai video gen can use any sd model, the thing is though that it's terribly hard to configure and very long to render. If you are not familiar with SD and its UIs you have slim chances of setting it up. Start with images first

[gilded dune]

https://www.youtube.com/watch?v=KBKQ7iQShHg

[twilit nova]

Thanks for your response.. I am familiar with Stable Diffusion. I am asking because I am not sure how much of these videos have been generated by a generative model and how much is the result of other post processing steps.

[twilit nova]

For example this video. How was it generated?

[long halo]

The T-Rex animation by omnison looks entirely done with automatic +sd to me

[twilit nova]

Just to give a background information, I am a student and I am watching AI generated videos to find out how close they are to real videos

[twilit nova]

Thanks

[twilit nova]

These as well.. They seem to be AI generated

[long halo]

SD stuff is very far currently

There are some corporate vid gens, which can make real videos

But they are terribly expensive

[twilit nova]

I know.. But I think the science is very close to generate very realistic videos. If this video for example is amazing if it was generated by Stable Diffusion

[elder hatch]

yeah stable video difffusion

[surreal tiger]

Sorry if this is a dumb question but when rendering img2img with a 1920x1080 batch what should I resize the outputs too? I have an I7 w/ 3090TI and my ETA was over 3 days. I interrupted that and then went to 1280x720 and it dropped to about 18hrs give or take. I am attempting to convert a video to AI and the original video is 1920x1080 30fps.

[clever comet]

For example, I use a workflow txt2img with US for the img, it then goes img2vid(svd1.1) and you may have to do several iterations, but you can get some good 5 second videos. It is the camera control like you see in Pika and RW that isn't quite there. Good luck!!

[normal veldt]

I used my dinosaur image and made 3 queries to Stable Diffusion image to video. For the second and third video queries I used the last frame of the previously generated video. I stiched it all together with pytho nafterwards.

[soft stratus]

This works.

[mossy anvil]

给我生成一个沙滩上的女人

[zinc yoke]

#1047610792226340935

[mossy anvil]

#1080946152318443610 Become a woman on the beach

[zinc yoke]

...

[mossy anvil]

How to use this software?

[zinc yoke]

install it locally or use online services

[mossy anvil]

It feels so complicated to use this thing in China

[last nimbus]

anyone tried and succeeded on training the svd model?

[gleaming parcel]

为什么？住在中国跟stable diffusion的使用有什么关系

[timber flower]

不会吧，有一大堆启动器，用启动器连环境都不用管，都给配好了

[somber onyx]

请问有用sd填户型图的吗？

网上下了一些模型填完 不是很理想

自己训练的也是不是很理想 没啥心的

[hard gyro]

https://www.youtube.com/watch?v=kks9RR1aLWE

whats this and how do i get it to work ?

its chinese i think

https://www.youtube.com/watch?v=5Tz1O1vZWLk

[zinc yoke]

seems like it s all explained on their github ( linked in description ) https://github.com/ali-vilab/VGen

[jolly flower]

really clean, realllllly smooth!! Any details you can share?

[shrewd crown]

Yeah of course, I've got a comfy workflow that does an XY plot for the video CFG scales, It's fed by a prompt, and had to play around with the movement scales to get just enough subtle movement.

[noble wagon]

https://youtu.be/ckvomWJoW_4

audiocraft, mixxx, scascade, svd, blender and deforum

[hardy sail]

how to use its software？

[unborn trout]

try https://voyager1.ai

Hey everyone! I made this using stablity.ai sdk
https://voyager1.ai

[tawny star]

https://youtu.be/Vc2hvp3P6ZA?feature=shared

[surreal plank]

/help

[rustic quail]

create

[round jasper]

how long does video generation via the API take?

[clever comet]

Really well done cars! For the SVD channel having each as a video moving would be awesome to see!!

[clever comet]

It's all going to depend on the hardware. Mine take about 3 minutes locally. But with api and the cloud could be the same or more/less time. All depends on what you are using.

[south wave]

https://www.youtube.com/watch?v=xfF69OTaJi4

[hard badge]

Hey, I have a conceptual question about SVD. We all know the hype around Sora making long form videos (60 seconds), whereas SVD can only generate a few seconds.

But what happens if you take the last output frame of an SVD generation and stick it back in as the input frame? Could you do this several times to generate a single, relatively cohesive shot that is 15 seconds long or more? This might not be a novel idea at all, if there are ComfyUI workflows or generations that do this already, feel free to drop links, I'd like to take a look.

I currently do not have the compute to run SVD locally so I cannot try this myself.

[robust spade]

Found this, just the generation no workflow- https://www.reddit.com/r/StableDiffusion/comments/183irwx/svd_over_24_frames/
Thats a really interesting idea. I guess this guy got a python script to grab the last frame and import it back into the original workflow. Ive tried a couple high frame animatediffs and some distinct images seem to last for longer than 4s pending framerate but always end up turning into morphs lol. I'm by no means an expert on svd/animatediff tho.

[hard badge]

Thank you for the pointer. As expected, you can see the degradation fairly quick, like ice cream melting on a hot summer day.

A follow up question, is it documented whether SVD can be conditioned on more than one input frame, to help guide the generation? (e.g. start and end frame, or start and middle frame)

[round jasper]

Thanks - I actually just saw on their website "2 minutes or less", which tracks

[ornate kraken]

https://www.youtube.com/watch?v=69f0PhWAZCg

[main mist]

Hello guys, when using SVD (in Forge, the inbuilt SVD tab), the cmd windows shows every time that SD loads the basic img2vid model, even if I select the sdxl1.1 one. Is this a bug or am I doing something wrong ?

[south wave]

https://www.youtube.com/watch?v=xfF69OTaJi4

Project files, and configs available!

[red hare]

Creative Ideas: The Adventures of Kidney Crusader, Superhero

**Story Overview.
In a fantasy world called Kidney Town, the inhabitants suffer a surreal plight due to kidney problems. Then, a superhero named Kidney Crusader appears with superpowers from MediCare™ Kidney Boost Drops, which he uses to save the residents from exhaustion and kidney stones.

Scene Setup (Scene Setup).

1. **Weary Daily Life: Residents move slowly like turtles or struggle like mountains without Kidney Power. 2.
2. **Kidney Stone Crisis: The residents are chased by a giant "Stone Goliath" and show extreme pain and panic. 3.
3. **Kidney Crusader's Arrival: Kidney Crusader makes his dramatic debut with exaggerated movements and special effects, wielding Nurbini™ Kidney Boost Drops as his "Power Source". The Nurbini™ Kidney Boost Drops serve as his "Power Source".
4. **Rescue Mission: Kidney Man uses his Power Source to unleash a powerful Kidney Blast to defeat the Stone Monster and revitalize the residents.

**Video Shots: Shot 1 (Shot 1): Kidney Man uses his Energy Source to unleash the powerful Kidney Blast to revitalize the Stones.

• Shot 1: The inhabitants, without Kidney Power, move slowly and display exaggerated expressions of exhaustion and discomfort.
• Shot 2: Residents panic as the Lumpy Monster wreaks havoc on the city in hilarious fashion.
• Shot 3: Kidney Crusader appears in slow motion with exaggerated special effects and an energetic superhero theme song playing in the background.
• Shot 4: Kidney Crusader uses Nurbini™ Kidney Boost Drops to unleash a colorful "Kidney Blast" that turns the stone monster into harmless rocks.
• Shot 5: The inhabitants are revitalized and begin to dance in energetic celebration as Kidney Crusader strikes a heroic pose.

Keywords.
#KidneyCrusader
#KidneyPower
#StoneGoliath
#NurbiniKidneyBoost
#SuperheroKidneys
#KidneyHealth
#NaturalRemedy
#EnergyRecharge
#TirelessVictory
#StoneBuster
#KidneyHealthChallenge
#HealthyLifestyle

[hushed tartan]

https://tenor.com/bLstt.gif

[stoic field]

https://tenor.com/i2vJFJF5Yb9.gif

/girl

https://tenor.com/view/puppy-gif-3427054701624599346

/StoneGoliath

[rough garnet]

Hola

[foggy needle]

draw a flower

[fast hemlock]

imagine prompt:A close-up image of a vibrant, colorful Jungle Juice Popper bottle surrounded by dense tropical foliage and exotic fruits, under a mysterious, moonlit jungle canopy. --v 6 --ar 16:9

[flat mulch]

animation, cinematic, 8k, 3d, high quality

[ruby charm]

even if it did work why wouldn’t you specify what kind of animation/subject

[trail atlas]

He get's it too

Greetings xD

[eager rapids]

/

[violet grove]

Video of the brand Defacto store with the logo

[violet grove]

The phrase coming soon with summer colors and a colorful golden rose in background

[violet grove]

coming soon

[violet grove]

coming soon

[violet grove]

coming soon logo

[ruby charm]

what…

[ruby charm]

what.

[frozen onyx]

Good afternoon, my name is Maxim. I'm from Russia. For a school project, I created a short cartoon using neural networks only. Please take a look and support. https://youtu.be/ruANV24h0Dw?si=Wr7Nbyo0Wr2_QB3-

[dim kraken]

#🏞｜general-with-images sunny lady

[delicate token]

What is the ideal settings to get smooth video?

[bleak cipher]

https://youtube.com/watch?v=MJaQEaeSLYE&si=vAvaywAIUDanCUkw

https://youtube.com/watch?v=miuy7xXv7Ps&si=q1VBZTvcrHtTfxUU

[grand siren]

Are there any comfy nodes for vid 3d yet

[mellow aspen]

https://tenor.com/YYtc.gif

一个大船在海上航行

A large ship is sailing at sea

[swift quail]

A large ship is sailing at sea

[errant cloak]

highly detailed and realistic 3D image depicting a 28 years old boy with wearable glasses joyful celebrating the Holi with lord Krishna in the vibrant setting of mathura Vrindavan. The boy is wearing a t-shirt
on which “Akshay” name is written, their faces are visible, the “Happy Holi” are elegantly scripted on the top of the sky. The background features the picturesque landscapes of mathura Vrindavan, complete with colourful buildings and lush greenery. Children are depicted playing Holi.

[gritty oar]

https://tenor.com/pc6HQ3hlWnK.gif

[south ore]

highly detailed and realistic 3D image depicting a 28 years old boy with wearable glasses joyful celebrating the Holi with lord Krishna in the vibrant setting of mathura Vrindavan. The boy is wearing a t-shirt

[echo pagoda]

low strain Li-rich cathode material

[bleak cipher]

https://youtube.com/watch?v=MJaQEaeSLYE&si=xA7QsUiPvRYiQhTy

[ruby charm]

there is no bot here

[topaz cosmos]

Given that SVD's initial image resolution is 576x1024 (or flipped), which is not an aspect ratio that SDXL images are trained on (closest is 768x1344), do you guys just crop the SDXL generate image before feeding it to SVD, or do you just say "close enough" and feed the 768x1344 (or 1344x768) image?

[bleak cipher]

https://youtube.com/watch?v=MJaQEaeSLYE&si=zdr9TgQ7pvwqVp30

[bronze aurora]

man

#🏞｜general-with-images man

[noble forge]

Friends, can anyone help me make an image, I don't know how to use it

[trail atlas]

did you know that you can combine infinite interpolators? just fast mode off, scale factor 4, and half the final output, i have 320 frames 768x576 in 160 seconds with 15 steps. 9 frames per second base for 54 frames. funny, cause the interpolator does even fill the gabs very vell.

[pale sapphire]

coming soon

[trail atlas]

you can text to video conditioning combine with the interpolator too to doubble the frames

instandly xD

and k-sampler rework them all

[noble forge]

Can someone answer my question? Maybe what you're saying is too deep for me to understand

[lucid canyon]

Can you explain more?

https://tenor.com/view/iq-low-dumb-stupidity-stupid-gif-24161342

[shut turret]

火影忍者

[noble forge]

@lucid canyonHow do I need to make the images I need

Sorry, I don't know anything about computers and I don't use English very well. In fact, I have to use a translator to communicate with you

[lucid canyon]

https://tenor.com/view/subradio-same-gif-22266555

[noble forge]

Can you show me how to use it, my friend

@lucid canyon

[lucid canyon]

https://www.youtube.com/watch?v=J4yOcyOafH0

[noble forge]

@lucid canyonFriend, do you have Chinese subtitles? I need your help

[lucid canyon]

https://github.com/ZHO-ZHO-ZHO

https://tenor.com/view/who-me-huh-what-wait-gif-25560426

[noble forge]

谢谢你

[noble forge]

@lucid canyonWhere are you from, my friend?

[tawdry parrot]

https://tenor.com/bEpef.gif

whats up

[silent hinge]

swimming fish

[wispy garnet]

all made by Japan-based AI creators

https://x.com/aicuai/status/1770468416740303174?s=46&t=jgQwk-8DSR4rWrfqK70Ylw

AICU Inc. (@AICUai) on X

All made by Japan-Based AI creators! The creative AI-driven reel made by #AICU Cooperating Creators. @8co28
@marimosphere
@Misositaworks

@ShojiHikaru @hane_desu @koji276
@ChiyamaKotone @o_ob @AICUai
@StabilityAI
AI Creators Union Demo Reel 2024 Summer https://t.co/vrxCboOqiD

Twitter · Mar 20, 2024, 3:12 PM

[marble yarrow]

"On a sunny day, the little duck and its mother are playing by the lake, enjoying the tranquility and joy of nature. The little duck is full of anticipation for the play that is about to begin."

[bleak cipher]

https://youtube.com/watch?v=5pnoRk_nN3E&si=vVe94_QI_00VKkH1

[ruby charm]

what?

[hard gyro]

My toaster had an existential crises third time this month.

[civic notch]

dog movie

[civic notch]

😆

cat movie too

[civic notch]

👍👍👍👍👍

[sturdy wave]

Our latest video was animated with SVD. Hope you guys enjoy! https://youtu.be/aHRTyOsPHBM

[sterile veldt]

nice work

[primal nimbus]

SV3D only generates orbital videos, no 3D mesh I think ?

[eager nimbus]

[sterile veldt]

well... that didnt work as planned

[hard gyro]

Haiper vs SDV

[pulsar finch]

https://youtu.be/HWCmO800YLQ

[iron nacelle]

@sturdy wave star

[flat whale]

bots offline #1047610792226340935 and you just pinged an innocent person

[crystal badge]

This stuff is dope

[sonic ravine]

Dog

[flat whale]

#1047610792226340935

[sonic ravine]

dog ¯_(ツ)_/¯

[flat whale]

#1047610792226340935

[tulip osprey]

😂 "Idc generate dog"

[flat whale]

stop posting about music for sleeping

[jagged tangle]

stop posting about music for sleeping

[flat whale]

butts.

[flat whale]

Can we just ban this dude from posting random YouTube links of music for relaxing

well I guess this is a form of promoting?

And does this count as copypastaing? Idk

At least it's annoying to me

[silent hinge]

HDR photo of woman, dark-brunette hair, light-blue eyes, (f_stop 5.6), (focal_length 28.0), f/5.6, 28mm focal length, sitting in bed, contemplating, sunset, romantic, warm, revealing clothes, alluring, white sheets . High dynamic range, vivid, rich details, clear shadows and highlights, realistic, intense, enhanced contrast, highly detailed

[soft stratus]

So how many channels did you need to post this on before you realized you were being an idiot?

[silent hinge]

Hey guys, what are the recommended settings for anime images? I tried this https://education.civitai.com/quickstart-guide-to-stable-video-diffusion/ and the image was messing up really bad

[sterile veldt]

Well… it says s on that link you just shared

But you are limited to frames

(Using the SVD workflow)

Can you share what you mean by “messing up real bad” ?

[silent hinge]

Before I do that, maybe it was because I changed the resolution from 1024w 576h to 1024h 576w, could that be?

[sterile veldt]

Couldn’t tell ya. No idea what you are even using as an interface

[silent hinge]

I'm using sd-webui-forge

[sterile veldt]

lol

Well that’s interesting.

Does it do the same with a non animated ? Like a different subject?

I believe that model wants a very exact resolution to start troubleshooting.

Turned my motion bucket way down, and set it all at 25.

I walked away from SVD until I can figure out it I can batch it

I need to see the output again lol

Could you not… render to a batch? And overlap???

Cuz this 25 frames sucks lol

[fading pagoda]

Can this picture be made to move?

[grim wadi]

A Rubik's Cube, each side can be composed of multiple squares, which can be optional, pushed, pulled, and telescopic as a whole. I want a video that can do optional, push-pull, and telescopic actions.

[flat whale]

#1047610792226340935 Bots offline. You can run Stable Video Diffusion locally or use online services.

[cosmic cobalt]

Hey guys, is it possible to make a comfyui SVD workflow, and then use it as an api for inference by passing an image?

[slim nebula]

/sdjf

[flat whale]

#1047610792226340935 bots currently offline, you can run it locally or use online services like https://seaart.ai

[neat glade]

waterfall, lake, sun

[neat glade]

waterfall ,sun, lake, clouds sky

[soft stratus]

How long before you get the hint?

-->#1047610792226340935 <--

[neat glade]

/waterfall, sky, lake, clouds

[glacial orchid]

Here is the image you requested.

[ruby charm]

the bot is down

#1047610792226340935

[distant bone]

waterfall ,sun, lake, clouds sky

[ruby charm]

No.

[storm glacier]

High-rise buildings, glass curtain walls, glass connection claws

[exotic badge]

A dragon came out of the water.

[ruby charm]

No

[ruby charm]

No

Wow the lack of awareness

Do y’all not see the first or second message you see

[outer lodge]

Today, we are releasing Stable Video 3D, a generative model based on Stable Video Diffusion. This new model advances the field of 3D technology, delivering greatly improved quality and multi-view.

[keen perch]

waterfall, sky, lake, clouds

[late oasis]

What is even going on here can you not read?

[twin crane]

https://tenor.com/5vZh.gif

[random bison]

amazing

[solar turtle]

黎明前的宁静被打破，天空中的星星逐渐隐去，一轮红日缓缓升起，带来第一缕温暖的阳光。打工人的房间内，闹钟的铃声在静谧的空气中显得格外刺耳。他从混乱的梦境中惊醒，睡眼朦胧，心中充满了对即将开始的一天的期待与不安，最好的质量，8k,生成一幅图片

[winged drum]

made with SVD and Haiper.ai : https://www.instagram.com/p/C5UM5a5IWnA/

cypher_wraith

001_CYPHER_04042024_01_FALLEN_ANGEL
#artificialintelligence #aiart #aifashion #generativeart #stabilityai #aivideo

Instagram

[bold rampart]

@neat glade high wind on dubai

[short condor]

https://youtu.be/mmFhyCy4sco

[glacial orchid]

Here is the image you requested.

[blissful cobalt]

a chinese dragon is swiming

[supple basalt]

小奶猫

[zenith ivy]

Generate cartoon pictures. In a warm town, there lived a cute little cat. The kitten has big eyes and striped black and white hair. There was also a little blue bird, a little blue bird, full of feathers, and big eyes.
The picture shows a cat lounging in a cozy bed, outside the sun is shining, but the cat is indifferent, next to the alarm clock indicating the time to be late. His mother looked anxiously at the watch at the door of the room, her expression anxious. By contrasting the warmth of the interior with the vitality of the exterior, this picture vividly depicts the dragging habits of the kitten.
The kitten's face is beaming with contentment and pride as his mother cat and little bluebird cheer him on. The picture not only shows the success of the little cat, but also conveys the positive results brought by advance preparation. Return seed value

/prompt Generate cartoon pictures. In a warm town, there lived a cute little cat. The kitten has big eyes and striped black and white hair. There was also a little blue bird, a little blue bird, full of feathers, and big eyes.
The picture shows a cat lounging in a cozy bed, outside the sun is shining, but the cat is indifferent, next to the alarm clock indicating the time to be late. His mother looked anxiously at the watch at the door of the room, her expression anxious. By contrasting the warmth of the interior with the vitality of the exterior, this picture vividly depicts the dragging habits of the kitten.
The kitten's face is beaming with contentment and pride as his mother cat and little bluebird cheer him on. The picture not only shows the success of the little cat, but also conveys the positive results brought by advance preparation. Return seed value

[frozen agate]

here is the quotation "game of survival, without compromising on, discovering one's true self" generate an image using red, orange and white text gradient colors respectively with suitable background color.

[slender pond]

a painting about a princess and a prince

[cerulean fox]

I made the most off the wall stable cascade and SVD workflow for multi subjects lol. https://civitai.com/models/385959/geekyghost-3-subject-cascade-and-svd-everything-but-the-kitchen-sink-workflow

[lilac cargo]

/prompte

[glacial orchid]

Here is the image you requested.

[shadow bay]

A simple design for a credit card targeting ladies with an Egyptian touch. It has to be simple

[glacial orchid]

Here is the image you requested.

[late oasis]

Here is the image you requested

[glacial orchid]

Beep boop! Prompt compliance error detected.

[round prism]

??

Is it supposed to be "Wrong answers only" ?

[dense galleon]

i'm goign to ask a question because:
A - I feel like an idiot
B - I can't seem to get it to work
C - someone has to know a 'simpler solution'

So i followed guide of https://education.civitai.com/quickstart-guide-to-stable-video-diffusion/

Still can't get it to work. got a 4090 - wanting to create short animated images (can be 1-2 sec could be 15 sec+ don't care).

Got stable difusion working but can't seem to get any of the other stuff to actually create animate work.
I've tried youtube guides (None worked)
Tried comfyui - Got it installed, but can't get the animated side working (shows a flow chart, won't import image - basically just geting rear kicked)

Anyone willing to help walk me through getting this setup?

Seriously I'm willing to tip a few bucks via paypal at this point

[chilly lintel]

/A cartoon hip hop rapper wearing jewelry performing in a dark cyberpunk background.

[teal arch]

Tried Stable Diffusion Forge? It's already integrated you only have to download the model ...

[normal basin]

Here is the image you requested.

[dense galleon]

Havent. Ill try. Ty

[teal arch]

It's limited to 25 Frames ... so FPS will set the video length ...

[dense galleon]

Yah im not looking at huge stuff. Just small movements for now. Trying to build stuff for tiktok for my books.

Any movement better than non for catching eyes

[teal arch]

Life of an artist

[obsidian glacier]

Life of an artist

[eager mural]

/eid mubarak card

[stiff lily]

do

https://tenor.com/8rAV.gif

[karmic schooner]

Streaming t2v is avaible for longer video generation with animatediff or svd

[cerulean fox]

Adding a bit more info about the nodes for this workflow since I combined some and use some lesser used nodes. https://civitai.com/models/388864/geekyghost-aio-workflow

[red vale]

生成一个Web3的视频

[daring comet]

@soft stratus 座椅

生成一把座椅

[neat bone]

A dog is running towards an old man in Memphis style with 4k image quality

[azure remnant]

When's the next video model dropping? I know devs read this, if you know just let us know. I will accept "no clue' as an answer lol.

[restive mist]

||draw dog||

[late oasis]

Here is the video you requested

[cerulean fox]

cascade and svd

[lilac barn]

any idea if zero123 works on MAC?

[azure remnant]

the best SVD / Suno video ever produced
https://youtu.be/40B6JVKGV_4?si=Mk5nOGa-U984WsV0

[wanton stream]

hello guys, could you give me some pointers? i need to find out how these two examples were made, deforum? animatediff?

https://www.instagram.com/p/C5QGkxhKV2s/
https://www.instagram.com/p/CvXrI-2o8fo/

michelin

No matter the weather, the MICHELIN CrossClimate 2 is developed to keep you moving ☀️🌧️🌨️❄️ ⁣
Discover more in bio ⁣
⁣
#Michelin #AllSeason #KilometerStory #MICHELINCrossClimate2 #UnexpectedMotion⁣

Likes

496

Instagram

evolutionofit

75 different BMW models in 21 second

#bmw #bmwmotorsport #bmwlife #car #cars #transformation #evolution #timelapse #fastforward #auto #viral #fyp #foryou #reels #bmwlovers #carlovers

Likes

6543457

Instagram

[ruby charm]

No

[glacial orchid]

Here is the image you requested.

[glacial orchid]

Here is the image you requested.

[acoustic wagon]

#🏞｜general-with-images Life of an artist

[glacial orchid]

Here is the image you requested.

[short pilot]

A handsome man in ancient costume drinks in a pavilion by the lake, and a beautiful woman in ancient costume plays in the water on a small boat on the lake

⁠🏞｜general-with-images A handsome man in ancient costume drinks in a pavilion by the lake, and a beautiful woman in ancient costume plays in the water on a small boat on the lake

#🏞｜general-with-images A handsome man in ancient costume drinks in a pavilion by the lake, and a beautiful woman in ancient costume plays in the water on a small boat on the lake

A handsome man in ancient costume drinks in a pavilion by the lake, and a beautiful woman in ancient costume plays in the water on a small boat on the lake

#🏞｜general-with-images A handsome man in ancient costume drinks in a pavilion by the lake, and a beautiful woman in ancient costume plays in the water on a small boat on the lake

[ruby charm]

no

[ruby charm]

no

[ruby charm]

no

[ruby charm]

nope

[ruby charm]

still no

[late oasis]

Here is the image you requested

[hot solstice]

Im working on a trailer and need to recreate the following 1080p video in cartoon/anime style. What would be the best way to do so ?

[teal arch]

DeForum you can use a Video as Input ... but never tried this way ... You can see the technology in videos of DJ Sickkick for Example ...

[teal arch]

https://www.youtube.com/watch?v=bicPayZDI60 (example)

[golden arrow]

how did you do that? impressive

[teal arch]

Some SVD Videos improved and cut together ...

[golden arrow]

am I being stupid.... whats SVD?

[teal arch]

Stable Video Diffusion. Kinda IMG2Video ... problem is: you don't really have control over the video generation ...

[golden arrow]

ok. where would you recommend I start. only want to create short 5 second movement

[teal arch]

SVD has a limit of 25 frames ... using 6 frames per second you will get around 4 seconds ... there are ways to improve quality afterwards. But I only know an expensive commercial way cause a friend is camera man and cutter.

Positive point: Control over start image. Negative: you don't really know how A.I. will animate it ...

[golden arrow]

does it take a lot of practice to pick it up?

[teal arch]

After a while you will learn how A.I. could understand a picture and animate it. But you only give a picture, only a few settings and get the video

[golden arrow]

ok. so not too many prompts?

[teal arch]

Leonardo.AI has this technology for example. They give a few tokens every day but never used it ....

Only your one prompt for the start picture ... and you can generate pictures till you have a good one and give it to SVD

[golden arrow]

pretty sure Run Diffusion has it but thought it would be too complicated and time consuming

[teal arch]

I'd think it's the easiest way. I have linked a video about Deforum above ... another technology that can use Video input + prompt ... but way more settings ...

[golden arrow]

ok. thanks for the advice. will check it out

[teal arch]

Good luck and enjoy the experience!

[golden arrow]

will do 👍

[undone granite]

Hello friends. please tell me where stable diffusion has gone and when it will be possible to generate it for free again, or tell me where it is possible now to generate photos and videos as before? Thanks in advance

[silent hinge]

What are some good settings to improve the quality of turnarounds with Stable Video 3D SV3D_u? I'm getting some good results but often there's a glitch near the end of the sequence. Any pointers to help me out?

[late oasis]

You got to download it

[ruby charm]

No.

[ruby charm]

The bot is down

[undone granite]

And what works? Where are the people now? I have a weak laptop

[hard gyro]

Musk Glaze

The bulletproof luxury car.

Where you retire your grandma in style.

Only two classes left. The have nots and the have debts.

[ruby charm]

Arby’s, we have the meats

[granite crypt]

If your program loses the generation id for a video, is there an endpoint one can use to fetch a list of your past generation ids?

[steady adder]

can I use a text prompt with svd?

[stark reef]

https://www.youtube.com/watch?v=orzLvvF8Bbs

[scarlet idol]

can we provide a duration to the video that is generated from image with svd

[umbral loom]

is stable video down?

[umbral loom]

nope, turns out my image wasn't the right size - sorry everyone

[solar turtle]

#▶｜stable-video-diffusion 生成一个机械哪吒大脑天空的视频

[peak nest]

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

[analog kiln]

how to generate anime vidoes?

[ruby charm]

#1047610792226340935 the bot is down

[past mesa]

[hard gyro]

The bot is on a bathroom break.

[peak yew]

阅读以下链接并生成一款新的狗爪板：要求跟链接里的材质、形状、功能都相似：https://www.amazon.com/FL-Falatier-Scratching-Multi-Functional-Alternative/dp/B0C6F6K5M2/ref=pd_vtp_d_sccl_2_7/141-8792912-4661849?pd_rd_w=fpzJr&content-id=amzn1.sym.1a17ce14-45e3-4517-a7b0-2d2405e8877f&pf_rd_p=1a17ce14-45e3-4517-a7b0-2d2405e8877f&pf_rd_r=FRK486QBH94DMYXFK0EV&pd_rd_wg=hJxm4&pd_rd_r=b413626f-1602-48d3-ba22-6bb94ef1a45d&pd_rd_i=B0C6F6K5M2&th=1

[silent hinge]

我不会啊

[sharp kindle]

Hey all, I'm super new to all of this and I'm hoping I can get some pointers here. I've followed along with a few tutorials and have spent the week trying to learn how to use comfyui and what it's capable of, now I'm trying to branch out and work on my own project. The only problem is...I'm struggling.

My goal is to take an old video, roughly 30seconds long and turn that into an animation.

This short video has 8 scenes.
2 of them contain 2 actors.
2 other scenes have out of focus actors cross the foreground.

Obviously I would like to keep a consistent character model throughout the video, this seemed easy but in one scene the character turns around and in general the motions seem more complex than the dancing TikTok tutorials that are all over YouTube when I look for a solution.

Using canny to map the characters positions onto the rainbow skeleton also only seems to partially work, I'm assuming this is a combination of foreground interference and actor motions. Should I just...create custom skeletons in paint for the unsatisfactory frames?

I'd love suggestions, advice, or links to more helpful YouTube videos than I've been able to find.

[outer light]

can someone please share some good API service for text to video generation ?

[pastel thistle]

a 30s clip at 60fps is 1800 frames. you'll never achieve temporal coherence with stable diffusion over that many frames. for a first project you should have way way lower ambitions, like maybe 2s video in a loop

[eternal wraith]

A digital illustration of a steampunk library with clockwork machines, 4k, detailed, trending in artstation, fantasy vivid colors

[elder escarp]

https://tenor.com/qaxMAch6Pxl.gif

[muted depot]

Anyone know why Im getting this issue. I have a good graphics card and there aren't many better ones so its confusing to hear that its a VRAM shortage in my GPU. Could it be something else or am I wrong?

[meager cloud]

Yeah 8GB is on the low side for certain for anything more complex than image generation. You can try ComfyUI started with FP8 unet in arguments to load models at half precision which basically halves your VRAM usage at a cost of running stuff a bit slower than usual and some tiny decrease in quality, but not sure if it also works for video models or not

[muted depot]

ok thanks

[ruby charm]

how will 64 gigs do?

[cosmic lynx]

hello

a running cute dog ,4K,detailed,3D

[rapid willow]

@silent hinge To improve the final output quality, try lowering the crf to 4-8. As far as glitches are concerned, make sure the clip step setting in you network matches the clip step setting on the CivitAI model card. Some require 2 and others 1. Also leave context_stride at 1 and context_overlap at 4.

@meager cloud Do you know what the command would look like to start comfyUI in FP8 mode? I'm using the GPU bat on windows. How would I modify it?

.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build

[meager cloud]

Run it with --help, it lists all arguments, there are like 2 fp8 unet modes and fp8 VAE I think but unet is the more important one, I think the one with longer name was a tiny bit better but overall they're very similar

Can't check rn because I forgot to turn my PC on in the morning so I can't ssh into it to check remotely

[ruby charm]

No

[cosmic lynx]

A man stands on the city wall

[strong frigate]

https://www.youtube.com/watch?v=Vjkpb57eeOo

Decided to opt for low FPS as an aesthetic choice, especially because my RTX 2060 only has only 6 GB. So this will do for now.
Will upgrade when the 5xxx cards come out.

[slender sparrow]

Why do I feel a little scared?

[strong frigate]

Probably the uncanniness of it all 😛
Was kind of going for the vibe of those dreams, where nothing is wrong, yet everything also feels off.

[strong frigate]

awww he's so happy about all that destruction.
Truly a mad man, better get out of his way.

[thorn umbra]

A flying bird

[ornate briar]

this has a lot to say. a deep deep lore, a lot of emotions, an entire movie

truly epic 🔥

[ornate briar]

yup same...i think it is because of the concept of uncanny valley and the liminal spaces..kinda like backrooms yk

looking human like, really close but not accurate

[strong frigate]

https://youtu.be/R9LxpESFRlQ?feature=shared

[sharp kindle]

are there any good animation resources that aren't just thirst trap tiktok dances?

[frank oasis]

https://tenor.com/bC5BD.gif

[dawn berry]

Thought Id share this one again

One of my personal favs

[azure remnant]

Is it possible to influence the motion direction when going from txt 2 video? For example, if do a image to video using a motion bucket of 30 and it creates 24 frames panning right over an object, is it possible to take the last frame of the generated video and continue that motion from the video before?

[primal pecan]

On a bright summer morning, a Golden Retriever named Xiao Wang decided to race the sun. Xiao Wang believed he could run faster than the sun and aimed to cover all the paths around the village before the sun reached its zenith.

As the story begins, Xiao Wang starts at the eastern edge of the village just as the sun peeks above the horizon. Excitedly wagging his tail, he begins his sprint along the country paths. He runs swiftly, almost feeling the sun's heat creeping up behind him.

Xiao Wang dashes past creeks and through shaded lanes, pausing briefly at each spot to check the sun's position. With every glance showing the sun a bit higher, he pushes himself to run harder. The villagers, inspired by Xiao Wang's enthusiasm, stand at their doorsteps cheering him on.

Finally, Xiao Wang reaches the western end of the village, panting as he stands on a hill looking up at the sky. The sun has reached its highest point, and though Xiao Wang hasn't outpaced the sun, he feels joyful and proud for giving his all and enjoying the journey.

The village children run over to surround Xiao Wang, patting his head and back, offering him the warmest embraces. Although he didn't win the race against the sun, Xiao Wang won the love and laughter of the villagers, making this an unforgettable day for him.

[sharp kindle]

Is there a way to comfy to add in the start keyframe and the end keyframe and have it generate all the inbetweens?

[dry berry]

How can I use this robot

[pure sparrow]

is stable video down?

their website and the Api are not working

[livid idol]

#🏞｜general-with-images cat

[glacial orchid]

Here is the image you requested.

[abstract mantle]

Is SVD still viable, or everyone switch to animatediff?

[fair forum]

一个人在月亮上行走

A person walking on the moon.

[strong frigate]

https://www.youtube.com/watch?v=bNCLWK0OHHI

[ruby charm]

🚶‍♂️ 🌚

[runic gale]

hello~~anyone knows how to control generate video time?

the video always 1sec

context length not working

[white elbow]

is there an API to generate video from stable diffusion?

[silent hinge]

guys wait for 50 series or buy 4080 now/

?

for stable diffusion video

guys how many minutes it takes to convert 1 video into animation which is 1080p@60fps and 30 minutes long?

[strong frigate]

Wait for the 50 series.
Currently waiting for the 50 series myself.

[silent hinge]

ooh cool

u have 40 series?

thats why u waiting?

i have a trash lappy

4 core

igpu lol

so i nvevr seen desktop in my life

[strong frigate]

I'm using a pretty low card atm, but it's getting the job done.
Image generation is super speedy and video takes 2 minutes for 2 seconds of video lol
I'm using an RTX2060 6 gig.

was considering getting 40 series, but i'm have a good feeling about the 50 series.
so i'll just wait a lil bit longer.

https://www.youtube.com/watch?v=CKAG96f2exk

Done using an rtx2060 6 gig. which is a super cheap card.
u could also opt for a 12 gig 3060 for the time being, and upgrade your gpu once the 50 series comes out.

like I'm finding barely used RTX3060 12 gig cards for € 220
a good alternative while u wait.
just my two cents lol

[vague vortex]

are amd gpus no good at SD?

how much trouble are they to use? I want to use SD with DR and Blender - so, should I not even consider an AMD gpu?

I might use Linux sometimes, too - I'm looking at used 3090s, 4070 Ti Super and the 7900 xtx

[strong frigate]

I always use NVIDA so I have no idea how well it performs on AMD.
It should work on AMD, but I do think it will perform much better on NVIDA.
That's me speculating though.
https://www.reddit.com/r/StableDiffusion/comments/1826sf1/how_is_amd_doing_with_stable_diffusion_nowadays/
Judging by the comments here, people aren't really a fan of how it performs.

I think you'll be better off just getting a 3090 or a 4070 Ti super.

[silent hinge]

plz tell me how long it takes to convert 1 video 1080p@60fps 1 minute video into animation with rtx 4080

[vague vortex]

thanks for the reply - the question then is which one - since the 3090 is more power hungry and older? I have a Corsair rm850w - it's enough?

the 4070 ti super has less vram obviously - 16gb total - and is newer but will be more expensive - even if I find one used

[strong frigate]

Those 24 GB will absolutely work in your favor.
My irl friend has a 3090 and stable diffusion runs like a dream.
if you want a newer card, you might as well just wait a bit longer for the 5xxx series.

[vague vortex]

Why? Is the 50 series gonna have more vram on more of their cards?

[strong frigate]

These cards will be based on the Blackwell architecture.
This link explains more about the architecture https://www.nvidia.com/en-us/data-center/technologies/blackwell-architecture/
So technically, if you're going to upgrade for Ai. then don't go all out yet with your money.
Next gen Blackwell powered cards are around the corner.
Will it be mind-blowing or will it be a dud? time will tell.
For now I 'm still ok with my very modest 2060 6gb, so I'll wait a bit till those cards are out.

[lavish ruin]

hello

[analog kiln]

how to generate anime videos?

[teal arch]

Getting old ...

[tulip sedge]

[glacial orchid]

Here is the image you requested.

[arctic copper]

I am always curious what the prompt from the initial message was.

[sudden trench]

Ya'll wan't some of that positive Solarpunk future where artists paint reality with AI nanobot swarms?

https://www.youtube.com/watch?v=Qgy9u_9fqKE

Made with mixture of a custom Solarpunk SDXL model, Stable Video and Runway

[glacial orchid]

me too

[white elbow]

thatss cool

[white elbow]

anyone using an api w sd video?

[ruby charm]

this is not a video

fyi

lol

[woeful spade]

The moment I realize I put a photo in wrong chat 😂

[feral sundial]

Guys I when I add the mov2mov extension it doesnt show up. Can you help me

[fierce star]

Story Diffusion when?

[pure yew]

a penknife in hand

#🌠｜show-and-tell a penkinfe in hand

[cursive arch]

Hi everyone, how to make this kind of video (https://www.youtube.com/watch?v=94ALmuvtBNY)

[sharp zinc]

Creat a cat

#✍🏼｜rules-and-tos a cat

[grave karma]

I want to draw a video with bright and rich colors, including a woman wearing red ancient clothes riding a horse through the desert, the woman wearing a veil, covering half of her face, showing a sense of mystery, and the style needs to be three-dimensional and realistic

[charred cypress]

https://youtu.be/3g_pqOESeYs?si=9H0XLLvxyoEkcJeV

[weary prism]

#dream person doing meditation background with universe and on his body all spiritual chakra are rotating

person doing meditation background with universe and on his body all spiritual chakra are rotating id:browse

[fluid plaza]

@runic gale Hey:) Did you find a way to solve this? I have the same issue

[old cobalt]

#dream, red planet on orange sky as seen from an Alien Metropolis

So how does the video generation actually work via Discord

[karmic flare]

Hi, a few days ago I came across a link that shows how to make an animation with masks. There was a picture of a hand holding sand or a squid and with the marking of a mask they made things move. Unfortunately I can't find it again. Does anyone know and can send me the link?

[true sparrow]

i hope they release more versions of svd

[scenic basin]

this maybe? https://huggingface.co/spaces/AP123/IllusionDiffusion

[glad marsh]

https://github.com/alibaba/animate-anything

[hushed hinge]

Anyone have any upscaler workflows that they trust and/or have tested for video upscaling?

[sharp shale]

Tested this and works fine: https://openart.ai/workflows/macaque_keen_26/video-upscaler/imOvpl0wQQQf6stGKS0p

[true sparrow]

svd 2 when

[sour skiff]

Top

Another one version

[glad marsh]

soon

[dense galleon]

thats some legit nice stuff

[frosty raven]

After endless bad generations, I feel like I finally found the right settings and these are the results. Motion bucket 120, augmentation 0.07

[dense galleon]

i still cant get SD to work for me.

sadly its just time getting to spend working on it..

i mean i can make stuff but nothign ever looks good and when i try to animate the images never stay consistent

watched a lot of youtube vids trying to copy but still no luck

[silent hinge]

guys how many minutes or hours it takes to convert 1 minute 1920x1080p video @30 fps into anime?

[frosty raven]

I'm not aware that stable diffusion video can do such a thing.

[silent hinge]

i seen on youtbe

[silent hinge]

guys can any1 send me 1 minute video 1080p@30 fps of man /woman converted into anime thx plz

i already requested ciso i think he is busy

[silent hinge]

https://rightbrainai-1307270943.cos.accelerate.myqcloud.com/2024-05-17/22-02-ee2c1dae-7538-4634-883f-f944ef36c609.mp4

thats me!

https://rightbrainai-1307270943.cos.accelerate.myqcloud.com/2024-05-17/22-12-baa80654-e2d7-4160-a4af-66bd18f2996b.mp4

which 1 looks more anime 1st or 2nd?

[left fable]

second

[silent hinge]

can SD do this?

i did this on AI called lensgoai

can sd do video into anime locally?

[silent hinge]

any1 tried novel AI

/help

[opal kite]

#✍🏼｜rules-and-tos Soldier sleeping in watermelon truck.

Soldier sleeping in watermelon truck.

[north rampart]

I have a local SD Forge instal and I want to try to do some videos. I can see that Forge has a separate SVD tab, so I assume I only have to get sdv model and throw it in to get starter? Or is there more to it?

[hushed hinge]

Does anyone know if the WAS node suite nodes work on video as well? Simple things like 'Add Grain' would be useful but I keep getting errors when trying to run them through video workflows..

[worthy whale]

#▶｜stable-video-diffusion a guy sleeping under a tree in summer

[forest haven]

why can we create imahes anymore?

[crystal maple]

Borboleta

[hallow meadow]

ice

[hollow magnet]

what did you use for this?

[frosty raven]

It took an image I generated with ELLA and dropped it into stable diffusion video

[hollow magnet]

whats ELLA and where is stable diffiusion video?

[frosty raven]

https://ella-diffusion.github.io

Sdv you can get from stability.ai

You can generate both on here too https://discord.com/channels/1002292111942635562/1237461679286128730

[hollow magnet]

Stable Diffusion Video doesn't work too good when I use it, Krea.ai Video actually works better.

I have 3 video codes to share, see the DM if you want it.

I was going to just paste the codes in the Krea Discord but then remembered you.

I'll give you one then give the 2 to others.

[frosty raven]

thanks, i'm ok with these tools for now. I'm content with the images side of things for the most part. the svd is just a bonus.

[hollow magnet]

its free

if you haven't tried Krea.ai Video

I sent you a code

I gave the code to someone else, I have 1 more

[hollow magnet]

Krea.ai Video is quite exceptional

any Stability dev want a code? it's free but with that much money, I'm sure they tried it

[hollow magnet]

if you aren't going to use it, I need to give it to someone else

[shrewd dune]

/girl

[west oriole]

#▶｜stable-video-diffusion un hombre caminando en las calles de Tokyo en el 2077

[hushed dust]

The story of Exodus is an epic tale

[barren hound]

Has there been made motion models that can do longer than 16 contextual frames yet?

[sharp shale]

Would like to generate videos of moving objects (Ex: Car) captured from the moving camera (Ex: Drone) where the camera angle as well as elevation changes. Anyone explored this ?

[novel karma]

风力发电的场景

[dreamy coyote]

what are you using for generating? Those looks really good

[broken otter]

Make the person on the left move his hands. The person on the right is nodding.

Make the person on the left move his hands. The person on the right is nodding.

[lime horizon]

#▶｜stable-video-diffusion Blackhole ejecting mass

[broken axle]

HOW??? did you do this

[pallid helm]

I've seen many many examples of pixel perfect smooth animated videos. Like @frosty raven showed above. Nearly always that are some figure making a slow turn for something like 2 to 10 seconds. Other than "small range of motion turns" I've also seen high quality ?morphing into different things" animations going on for 1 minute. I suspect to produce the morphing thing takes ?hours? of GPU and a day(?) to fix up, tweak, edit, and polish.

Where might a lower quality but real-time interactive video creation play in this space? Something that doesn't require watching hours of Youtube videos to learn various techniques and tools and then still require many hours to create content. Something for a dummy like me.

[pallid helm]

Something like this 38 seconds thrown together which took 38 seconds to generate at 1024x1024.

Sorry, for some reason I can't hear the audio on this where I say:
bird and flowers
a clock
Tom Cruise
A mountain garden
An Asian women in a red dress

Zoom in, out, drag

to show a bit of its capability. 😦

[last folio]

/animated

/made a video

settings

[last folio]

/credit

[last folio]

hi

[last folio]

#artisan-faq

[sacred loom]

Once upon a time, in a peaceful fishing village, lived an old fisherman named Santiago. For many years, he hadn't caught a big fish, but he always believed that one day his luck would change.

how to use it？

[soft stratus]

#artisan-faq

Go there. Read.

[sacred loom]

thank u

[fading pagoda]

/made a video

/credit Three or five Chinese children running on the grass, backlighting, carrying backpacks, blue sky and white clouds, surrounded by large grasslands, trees in the distance, with an oil painted background. In a realistic and ultra detailed rendering style, the fine gloss, 8K

[south needle]

/credit

[daring light]

generate random video

/made a video
/credit Three or five Chinese children running on the grass, backlighting, carrying backpacks, blue sky and white clouds, surrounded by large grasslands, trees in the distance, with an oil painted background. In a realistic and ultra detailed rendering style, the fine gloss, 8K

[junior nymph]

/made video
/credit Girl with a gun in her hand and a phone in her hand

/credit

[lucid radish]

A lotus comes out of the water, with huge flowers blooming and several lotus leaves covered in water droplets on the surface.

A lotus comes out of the water, with huge flowers blooming and several lotus leaves covered in water droplets on the surface.

[turbid rampart]

what are you doing

"oh if typing random stuff to get generations didn't work the first 5 times it must work the 6th time randomly" seems to be the general idea

[shut sorrel]

Animal in the forest

[tiny wadi]

🤣🤣🤣

[hard gyro]

water and fire still looks funny

[turbid rampart]

HOLY SHIT WHERE

[neat zinc]

can stable video diffusion be conditioned on text as well as imgs?

it's conditioned on clip embeddings right? is this where the image conditioning happens or are conditioned images concatenated to the noise? thinking could replace the clip image encoder with the clip text encoder for this. would this work?

[cinder ember]

/made a viedo /credit Three or five Chinese children running on the grass, backlighting, carrying backpacks, blue sky and white clouds, surrounded by large grasslands, trees in the distance, with an oil painted background. In a realistic and ultra detailed rendering style, the fine gloss, 8K

[modern walrus]

is there a way to use animediff or something similar without an input video?

[bleak apex]

yes in comfyui, search the places that share workflows

[modern walrus]

I do not know how to use comfyui

I guess i should learn

[sour pumice]

That is impressive!!! How to do that?It's so stable!

[pine viper]

Does anyone happen to know what kind of software is used to achieve such movement video results? Is SVD used here? (reference video: https://www.tiktok.com/@moozoo_aicat/video/7372522628889972014 )

[next knot]

Create a video about a man looking at the scenery at the top of a mountain

[next knot]

"Could someone please tell me how to create a video here?"

[analog egret]

You create it in the #1237459938901491852 channels .

[soft stratus]

Is this a famous phrase? Who were you quoting?

[sterile mesa]

"Could someone please tell me how to create a video here?" - Winston Churchill

[hollow gyro]

cross-posting in case someone has an idea for this one #🤝｜tech-support message

[dusty harbor]

Haiper AI VS PikaLabs VS Stable Video Diffusion (S E A S O N S) #shorts
https://youtube.com/shorts/9zuPCRH5AA8?feature=share

https://www.instagram.com/reel/C71cAa1t06f/?igsh=eHRybHUxOXltaGp1

Instagram

[frank wave]

A blue 6inch tall dinosaur being crochet by human video is the humans hands above a desk crocheting with blue yarn a 6inch tall dinosaur

[hard gyro]

haiper is best

[median talon]

As far as I know there is no way to use a Lora with stable video is this true? Is there any community that is working on this that I could talk to?

[neat nebula]

Hey there.

[hard gyro]

https://www.youtube.com/watch?v=BTfLq-XkO0w

Im tellin u sora will be obsolete by the time it drops

[teal arch]

I don't think I've shown the final version here yet. A.I. Song and video. Never knew I could do something like that 😄 Nothing to buy here. Just have fun! https://youtu.be/vxLnqgTCloY?si=iB2ka63ipjGYN1g6

[astral crypt]

Oh my God ! Das ist ABSOLUT genial, Kompliment ! I'll link that to some friends and family in Germany. Excellent job with SVD 👏

[teal arch]

Thanks! I've had the fun idea, experience with Stable Diffusion and learned the rest while working on it 🙂

[astral crypt]

Hast du toll gemacht !

[vivid garden]

/describe

[rough musk]

/help

[astral crypt]

A video entirely SD-SVD made, right in time for SD3 release ! https://youtu.be/O3NzGSHjj4s

Turn your speakers on, the song "The Ballad of Comfy UI" is also AI made : Udio & Mistral

[alpine jolt]

create a cow riding a frog on a fantastical journey into the wild.

[ember nexus]

On a rainy day early morning, road filled with hazy fog and a paddy field with a scarecrow.a cop finds a corpse with blood splatter in the shirt and shocked with a confused and investigative mindset :

#1070829655638151208

[strong frigate]

https://www.youtube.com/watch?v=rxweJUsC0mE

[muted thicket]

anyone know of a way to export a sequence of images where vram gets cleared after each image export? im trying to animatediff at a higher resolution but batch rendering a bunch of large images crushes my 4090 card's max vram. I only get images to about 512 x 760 but no bigger.

[hollow sinew]

Create a video of a man running through a rainy city at night

[crisp badger]

How Create video?

[compact marsh]

Pink zebra

[iron token]

A man doing skydiving

[rancid belfry]

https://storage.cdn-luma.com/lit_lite_inference_im2vid_v1.0/1759a568-b1bc-4abe-9806-43b354a2c265/watermarked_video096e94dd44aab448aa0b4141a4f7f8865.mp4

https://storage.cdn-luma.com/lit_lite_inference_im2vid_v1.0/ee59eb62-7544-4f1f-bc42-1ef159e9e2df/watermarked_video00f64ca9c9fbb46b7b56a82dce36198a8.mp4

[solid cargo]

Turtles Love Slow - An AI Film Created with Stable Video Diffusion
https://youtu.be/HkHBpcixNNM
Music By DJ Rocswell, available now on all streaming platforms

[tidal pendant]

i havent messed with stable video diffusion in a long awile but i just put an image of sd3 into it

[wide thicket]

This is a luma video that I downloaded from somewhere. Can someone tell me where to use luma? I want to generate more according to my own ideas.

[fluid pulsar]

mutliple tools used together

[sand flicker]

Lumalabs.ai

[iron topaz]

https://lumalabs.ai/dream-machine

[velvet pendant]

how to promt videos here?