NEW LTXV2.3 EasyFlow - v1.7
Create one-shot videos up to 30 seconds long !*
*(depending on your setup)
Files choice for I2V or T2I:
Archive file is the standard v1.7 for I2V users, using image as T2I latent size.
Config file is the custom v1.7 for T2I users who want custom T2I latent sizes.
This version includes :
Better and Easier UI
Prompt saver
Prompt Enhancer by LTXV2
Low Res - Debug and Process section
2 or 3 stage generation
Choose model type (Safetensor or GGUF)
Control video slicer limit
Save seed with prompt
3 stages preset resolutions
Control first and second/third pass distillation
Exposed Sigmas
Added RTX VSR for Upscale
Other minor adjustments
New features / Minor changes
Safetensor and GGUF models
Safetensor and GGUF models/Vaes/Clips are all in the same subgraph now, activation is still the same using the fast group bypasser.
Control distillation on passes
Now you can control the strenght of the Distilled LoRAs on first and second/third pass. Second and third passes are using the same strenght, I'll change this in the futur if needed.
Other Minor Changer
I added RTX Video Super Resolution for uspcale. I highly recommend to run with only one upscale method to avoid exploding your RAM usage.
Now Sigmas are exposed at the bottom of their generation group. Feel free to modify them as you need.
Same system, better visibility. Just choose when the slicer start to cut the last second of the video.
About LTXV Prompt Enhancer:
If you are an advanced user, I highly suggest you to replace this enhancer by another nodes using LM Studio and ask prompt enhancements to uncensored LLM models.
Models and explanations inside the Workflow !
This workflow is under constant optimisation, expect to see an update at some point. Feel free to ask questions if needed.
Thank you for using this ComfyUI workflow. Please use it responsibly and respectfully.
This workflow was created for creative, technical, and personal experimentation — not for harm, harassment, or any form of abuse.
HAVE FUN 😁
NEW LTXV2.3 EasyFlow - v1.5
(Preview samples are not post processed)
Create one-shot videos up to 30 seconds long !*
*(depending on your setup)
This version includes :
Better and Easier UI
Prompt saver
Prompt Enhancer by LTXV2
TensorRT Upscaler
Low Res - Debug and Process section
2 or 3 stage generation
Choose model type (Safetensor or GGUF)
Control video slicer limit
Save seed with prompt
3 stages preset resolutions
New features / Major changes
Now you can run 2 or 3 stage generation.
2 Stage: This is the standard one in the v1.0 and official LTXV workflow. This one takes your image size and divide it by 2 to create the video latent size, generate first pass at low res, then second pass for the x2 upscale.
3 Stage: This is the new one. Like the 2 stage, we take your image size and divide by 4 to create the video latent, generating first pass at very low res, then second pass to upscale toj x2, and 3 pass to upscale to x4 total. This take a bit more time but the final output is more dynamic (also audio) than 2 stages mode.
The 3 stage mode uses your original resized image. However, at these lower resolutions, certain sweet spots of the model are preferable and can be activated with the button shown in the image below. The latent size will be automatically resized to predefined dimensions based on your image's aspect ratio (landscape, portrait, square).
⚠ Important note:
On the 3-stage mode, it is important to keep the seeds of the uscpale passes identical and fixed, otherwise you will have texture smoothing and blur on the final render.
Control video slicer limit
To avoid generating an unnecessary extra second on short generation times, you can now choose the duration at which this extra second is generated and cut off. Value is in seconds. This system is there because depending on the loRA or parameters, I have weird hallucinations or burns on the last frames.
(In this example, it will generate an extra second if the duration is equal or more than 20 seconds. Under this duration, there is no more extra second).
Models and explanations inside the Workflow !
This workflow is under constant optimisation, expect to see an update at some point. Feel free to ask questions if needed (use my discord link on my profil, it's easier for everyone).
Thank you for using this ComfyUI workflow. Please use it responsibly and respectfully.
This workflow was created for creative, technical, and personal experimentation — not for harm, harassment, or any form of abuse.
HAVE FUN 😁
LTXV2.3 EasyFlow - v1.0
Create one-shot videos up to 30 seconds long !*
*(depending on your setup)
This workflow is a humble modification of the basic LTXV2.3 template, adding a few tools and a better UI.
This version includes :
Better and Easier UI
Prompt saver
Prompt Enhancer by LTXV2
TensorRT Upscaler
Low Res - Debug and Process section
On demande - Alternative version with Grok Auto prompting (API KEY needed)
Models and explanations inside the Workflow !
This workflow is under constant optimisation, expect to see an update at some point. Feel free to ask questions if needed (use my discord link on my profil, it's easier for everyone).
Thank you for using this ComfyUI workflow. Please use it responsibly and respectfully.
This workflow was created for creative, technical, and personal experimentation — not for harm, harassment, or any form of abuse.
HAVE FUN 😁
Description
Features added:
Choose model type (Safetensor or GGUF)
Choose stage mode (Run in 2 or 3 stages)
Control video slicer limit
Save seed with prompt
FAQ
Comments (13)
This workflow works great. Had minimal custom node issues on a fresh ComfyUI/Stability Matrix install. Which is a deal breaker for me. If I have to fight with custom nodes that haven't been updated in 2 years I just delete the workflow.
Thank you for your feedback. Which custom nodes gave you trouble? I need to take a closer look to this, wich I didn't on this workflow because I updated everything a week before making it.
@thedrekt00 None of them gave me actual trouble. It's just always a dice roll if you'll be able to download them from within ComfyUI or not. In this case if I recall correctly everything went just fine. I can't remember 100% though. However looking at my download history in Firefox I don't have any repo zips in that history after downloading your workflow so I didn't download any custom nodes after grabbing this workflow. I probably should've phrased that better, but my brain was mush sorry for the confusion.
The only problem I hit was all the errors it throws from missing models because it does it for all nodes and doesn't exclude the disabled ones. So i had to weed through them one at a time to figure out the false positives. Which is only difficult because I am still very new to Comfy and it take me a minute to figure out what I need to download, and what I already have a version of. I'm really trying not to download 9 petabytes of models every time I try a new Workflow. 🤣 But all of that is a me problem.
@hardwire666 Oh no problem for the confusion, happy to hear that everything worked fine. And it's okay, we all been through this at some point, nothing better to learn than doing errors. Are you not using ComfyUI Manager for the custom nodes? If not, I highly suggest you to install it, it will change your life about downloading and installing custom nodes.
@thedrekt00 In the more recent versions of ComfyUI the as i understand it the manager was folded into ComfyUI and turned into the "extentions" button. When I first started with comfyui I lost a whole day trying to understand what "the manager" that every one mentions is. Also yes I do use it, however not all workflows are created equal and some use nodes that you can't find using it.
@hardwire666 They integrated the manager into Comfy? I hope we're talking about the same thing, because if it's the same "manager" accessed via the extension button, it's a load of crap, or at least it used to be. We are talking about the manager that lets you upgrade nodes, upgrade Comfy, install nodes with GitHub links, the model manager, and the restart button (ComfyUI Manager V3.39.2), right? You should have a blue "manager" button in the upright corner in Comfy (at least for my version).
Yes I understand, I downloaded a lot of Workflows with either too many custom nodes or removed nodes, since then I try to be careful, in my new Workflows, to have the maximum of accessible and simple nodes.
@thedrekt00 Yeah we're talking about the same manager.
@hardwire666 Yeah okay (I asked because I had someone saying the same but still using the old Extension manager, that's nice they've intagrated the manager in the new versions of ComfyUI, it will be easier for newcomers. In any case, thank you for the exchange and your feedback.
@thedrekt00 You're welcome!
can't find the toggle for T2V .... ?
@itasky At the bottom of the 2 or 3 pass subgraph. Next version will have a better UI for this option.
@K0DEX thanks, the one called "Value_2" i guess. Btw why not using the "power lora loader" that is more versatile and less basic
@itasky @itasky Shouldn't be displayed as "Value 2" but "Bypass I2V", maybe you have something doing a conflict with subgraph.
For the "power LoRA Loader" I don't remember what motivated my choice, I think that at this time I had conflicts with some important nodes I was using on other workflows and rgthree nodes, I was just tired having the "Power LoRA Loaders" not loading or doing a reset on every refresh. Now having a clean install without any conflicts after breaking my last Venv (🤫), I'll put them in the next version.
It will be integrated on the next version, adapted for the latest "10Eros" model, replacing the current loaders, coming through this week or next week-end, depending on my free time.