LTX Image2Video workflow with Camera Control and Flow Edit Upscaler. You can also easily control the amount of motion in the output video and avoid still frames.
____________________________________________________________________________________________________________________________
Upd 1.4
Udapted to LTXV 0.9.6 Dev & Dist models.
(Distilled model insanely fast - 5sec for a gen!)
____________________________________________________________________________________________________________________________
Upd 1.3
Udapted to LTXV 0.9.5
Reworked Upscaler - now it's based on FlowEdit. Removed FE-pass, as it was almost never used. Added optional SpeedBoost group (for those who has Triton, etc).
In the new version of LTXV, nothing much has changed in terms of quality, so the advice remains the same - search for seeds in lowres until we get the desired, then upscale.
____________________________________________________________________________________________________________________________
Added previous version (v1.2) based on LTXV 0.9.1. In my opinion this version works better. You can try it yourself and compare.
Description
LTX Image2Video workflow with RF-Inversion Upscaler and Camera Control (using standart and DepthFlow nodes). Also you can easily control amount of motion in your output video and avoid still frames.
Upd 1.3
Udapted to LTXV 0.9.5
Reworked Upscaler - now it's based on FlowEdit. Removed FE-pass, as it was almost never used. Added optional SpeedBoost group (for those who has Triton, etc).
In the new version, nothing much has changed in terms of quality, so the advice remains the same - search for seeds in lowres until we get the desired, then upscale.
FAQ
Comments (24)
hi. why am i getting these errors:
got prompt
Failed to validate prompt for output 734:
* FloatConstant 714:
- Required input is missing: value
Output will be ignored
Failed to validate prompt for output 735:
Output will be ignored
Failed to validate prompt for output 658:
* DepthflowMotionLinear 666:
- Required input is missing: high
* ImageResizeKJ 605:
- Required input is missing: width
* VHS_VideoCombine 93:
- Required input is missing: frame_rate
* Depthflow 672:
- Required input is missing: num_frames
* ImpactSwitch 692:
- Required input is missing: select
* LTXPerturbedAttention 80:
- Required input is missing: cfg
* BasicScheduler 313:
- Required input is missing: steps
* SamplerCustom 83:
- Required input is missing: noise_seed
Output will be ignored
Failed to validate prompt for output 93:
* (prompt):
- Required input is missing: frame_rate
Output will be ignored
Failed to validate prompt for output 383:
Output will be ignored
Failed to validate prompt for output 713:
Output will be ignored
Failed to validate prompt for output 740:
Output will be ignored
Failed to validate prompt for output 606:
Output will be ignored
Failed to validate prompt for output 739:
Output will be ignored
Comfyui -> Manager -> Install Missing Custom Nodes
@vbodnar there are no missing nodes there. it is not about the nodes but I do not know why somehow the set nodes and the get nodes just can't send the values
Hi, very nice thank you! But are you sure that the camera control actually works? I find it that is more or less random...
LTX is a black box. Camera control is work. You just need to find a right Seed) see rule number 1
@N0NSens Actually after some testing I have found that it works very nicely, but only with lower resolutions - I've tried 1280 as default first and the result is much worse even with 0.9.5 model... Thank you for the workflow, it is really nice!
Hey, I’ve been testing this workflow, but my videos are still super static. 😅 I’ve tried tweaking Seed, CFG, CRF, FRate, and Camera Steps, but nothing really changes.
Does this workflow only work properly with LTXV-DREAM-DIFFUSION-VIDEO-3B-V1.safetensors? I’ve been looking everywhere for a download link, but I can’t find it anymore.
interesting. Can you share any of the problematic images?
@N0NSens Thank you for your response. I used the default settings from the workflow, except for the model, which is LTX-Video-2B-v0.9.1. The image resolution I used is 768x512. Additionally, Camera Control, LTX Upscaler, and Flow Edit are disabled, resulting in no motion.
@N0NSens I also tried deleting all files in the custom nodes folder in case there was a conflict. I updated ComfyUI, ran your workflow, and installed the missing nodes, but the result is still no motion.
Well, without seeing the original picture, it's hard for me to answer. As you can see from the gallery here - all kinds of pictures are animated without any problems.
@N0NSens I just tried taking a screenshot of the images in the animation on this page. I adjusted the resolution, processed it in the workflow, lowered the CFG, increased the CRF and frame rate, but the result still has no motion. I will try again later.
Same here, I suspect LTXV-DREAM-DIFFUSION-VIDEO-3B-V1 is just good at creating movements
I finally succeeded in creating the video using this workflow. I decided to remove the Video Combine nodes and replace them with new ones. Thanks a lot! 😊
nicely structure workflow but I am getting static output regardless of inputs. It's not due to DreamLTXV - it works with the i2v workflow that was posted with the lora release. perhaps there are some nodes that are not passing params as expected?
Can you share your problematic image?
Try deleting the Video Combine nodes and replacing them with new ones. I failed before too, but now it’s working! Thanks, NONSens
@5a3 It works
@eakkawat Nice
@5a3 thanks that worked
I suffered for a week, but still did not achieve normal results. Reference images from the author's video
what to do with this?
I checked many images, no result
@sd_animation_art If you mean the result is static, try removing Video Combine and replacing it with a new one