Skip to main content

"The spirit of open source community is about sharing knowledge"

Q ≈ T + E
(Quality ≈ Time + Energy)

RESEARCH & DEVELOPMENT (2026)

(FREE WORKFLOWS)

To keep up to date with latest research, workflows, and content creation, visit my Patreon (free tier) or subscribe to the YouTube channel

Due to the constant time-consuming need to maintain this website with updates, I will henceforward share new and updated workflows only via the Patreon (free tier). They will be accessible to everyone, even non-members.

This site will host all my legacy workflows and you are welcome to download them, but I won't be updating this page as regularly with workflows from this point onward, only via the Patreon posts and via YouTube channel - Mark, 16th April 2026.

For 2025 research posts and workflows visit Research (2025)

Topics are presented below in alphabetical order, and not in the order videos have been made. To aid speedier page caching, videos have been hidden behind clickable grey windows.

ABOUT

The below is research begun in 2026 and includes free workflows. 2026 brings a change of focus toward content creation. All previous research can be found on the Research (2025) page and much of it is still viable.

To keep up to date with new research and content creation, join my Patreon (free tier) or subscribe to the YouTube channel

💻 MY 2026 HARDWARE & SOFTWARE

PRE-UPGRADE SETTINGS:

The above was the best settings I got to at the time, and then stayed on when I got there until March 2026 because "If it aint broke, don't fix it".

POST-UPGRADE SETTINGS:

A video detailing my experience can be found in the Install ComfyUI Portable section.

Date: 7th March 2026.

I have now upgraded the software ready to take advantage of LTX-2.3 and because ComfyUI was increasingly forcing us to have to. It was worth it. Note: my hardware remains the same.

For now I wont add more switches in while I test the baseline but I can say it has gone well and though some small issues it's also clear it has sped up certain tasks. I'll share more as I learn it. - Mark, 7th March 2026.

WHY ARE YOU USING SUCH LOWVRAM?

All my 2025 research was done on a regular home PC using OSS software and models. It was running 24/7 quite often, and is still going in 2026. I like the challenge, and if an AI task is going to take 15 mins I have time for doing something else. Also, the devs keep improving the code so we can use it.

As price hikes continue it is all the more reason to work with lowVRAM equipment. My current ethos is to work with affordable LowVRAM equipment to provide the best hope for other users who cannot afford expensive hardware.

Remember, most artists are poor and are not likely to see that change because of AI.

HOT TIPS FOR LOW VRAM: When using a workflow with previews in I recommend running it with the ComfyUI browser moved to an empty part of the screen. Tests on my setup have shown me that GPU "blap" is being caused on my lowVRAM by trying to watch video previews. This causes huge impact on speeds. Something to be aware of. I recommend process tracking your GPU constantly to know what it is doing.

HOT TIPS FOR LOW VRAM: In Firefox (not tested others) when running intensive workflows (esp effected is batch runs) open another browser tab and sit on that page instead of ComfyUI. I find this stops the "GPU Blapping" that adds time to my runs especially batch runs. I thought it was VRAM issue, or because videos auto play in the screen, but turns out something about ComfyUI being the focus window causes the problem. Weird, but try it.

back to top of page

BASE IMAGE PIPELINE

Due to the constant time-consuming need to maintain this website with updates, I will henceforward share new and updated workflows only via the Patreon (free tier). They will be accessible to everyone, even non-members. - Mark, 16th April 2026

QWEN IMAGE EDIT 2511, Z-IMAGE, SEEDVR2 (4K)

Base Image Pipeline (Qwen Image Edit 2511, Z-Image, SeedVR2 (4K)) - April 2026 video

Date Workflow 1: 20th January 2026.
Date: Workflows 2: 3rd April 2026.

About: This is my base image pipeline for getting from idea to base image. It may change as new tools develop and I will present them here when that happens.

This base image pipeline is good for character creation, camera angles for Lora training, realism, fixing plastic faces, fixing issues in shots, moving round shots, shot development from an idea to First Frame and Last Frame.

It is everything you need to be ready to start working on an idea and be ready to convert to video form.

This is the first time I have been able to get to 4K with ease, and I only bother because the process is so fast it makes it worth it.

The process outlined is fairly simple:

  1. QWEN IMAGE EDIT to create an image or adjust one.
  2. Z-IMAGE with low denoise to add realism.
  3. SeedVR2 to upscale to 4K.

I also talk about Krita and the ACLY plugin in the video, but only briefly. The ACLY plugin gives Krita access to ComfyUI models.
Links to Krita are in the video text or in Useful Software section at the bottom of this page.

Workflows 1: (This is now superceded by "Workflows 2" download below, but may have useful entries so will keep here) To download the various workflows shown in this video right click here and download the zip file. The workflows can be found in the json files and you can drop them into Comfyui and they will load up. It contains the following:

  1. MBEDIT - i2i-ZIMAGE_add-realism_vrs1.json
  2. MBEDIT - Lanpaint-Z Image Turbo Fun 2_1 InPainting 4.0.json
  3. MBEDIT - QWEN-2511-ref-image-restyle_vrs2.json
  4. MBEDIT - QWEN-Benji_image_edit_2511 (multiangles)-V2.json
  5. MBEDIT - SeedVR2_4K_image_upscale.json

The above are available in some form everywhere and don't contain anything new other than the models I use are Quantized. The difference is really in approach, and that is discussed in the video. I havent shared info on where to get models but notes in the video text explain that further. It is recommended you get the models that best suit your hardware setup.

Workflows 2: (Uploaded 3rd April 2026) To download my latest image pipeline workflow right click here and download the zip file. The workflows can be found in the json files and you can drop them into Comfyui and they will load up. It contains the following:

  1. MBEDIT - SeedVR2_4K_image_upscale.json
  2. MBEDIT - Zim-t2v-2pass_vrs5.json
  3. MBEDIT - i2i-zimage_add-realism_vrs7.json
  4. MBEDIT - Krita-QWEN-2511-swap-out_vrs15.json
  5. MBEDIT-QWEN_image_edit_2511-multi-camera-angles_vrs2.json

What's new in workflows 2: (Uploaded 3rd April 2026) The above contains some updates to the QWEN nodes, just use of multi camera angle for one shots and two loras; one for head swapping and one for blending in objects, both are very useful tools. But the real new hot thing is the zimage 2-sampler pass. I use it less in the i2i version but is best used in the t2v workflows for Z Image and provides great results. Though it still sometimes fails with hands and stuff. Generally the quality is much better and its still very fast. It works by making the structure very small in the first sampler (288 x 288 or equivalent size if using different aspect ratio) then upscaling x6 in latent space before passing it through the second sampler. Good trick. I am using this now in my basic image pipeline until further notice. I also tested Klein but I honestly do not see its value over QWEN 2511 that remains my go to with Z Image as the fix up, then SeedVR2 as the upscaler to 4K. - Mark, April 2026.

back to top of page

QWEN-KRITA WORKFLOW

QWEN & KRITA Workflow Base Image To Video (For Developing New Camera Angles)

Date: 6th March 2026.

About: If you are making base images for narrative-driven story-telling then you need consistency. For that, you are going to need some kind of image editing tool.

Krita is my go-to for this, mainly because it offers connection to ComfyUI and use of all the models via the ACLY plugin. Both are free and Open Source. You could use Adobe or Gimp or any decent image editing tool and the approach will be the same. This is about fixing up details in images that ComfyUI gets wrong.

The video shows an example of how I jump back and forth between QWEN and Krita to get a final base image looking right. You really can fix anything if you spend the time on it.

In the example I start with a shot in a space ship, use QWEN to turn the camera around 180 degrees. (This always has issues with all models and there are technical reasons for it. Even subscriptions models struggle with the perfect 180).

So next I go to Krita and a lot can be done there inpainting manualls. But more than that, with the ACLY plugin you can also export the exact workflow in Krita back out and load it up to ComfyUI. I often do this just to see how ACLY got it working. (I also do this for SDXL models where Krita is super fast for removing masked subjects).

In the video I showcase a blending "fusion" lora which will blend in a character even with white background still in place, and put it into a scene and relight it. And the bsf head swap lora, which also has a version for LTX and Klein 9b (I struggled to get it to run on my 3060 RTX for LTX, sadly).

Less than a mornings work and I get from idea to finished video clip of the captain stood on the bridge in front of a lift door, with the correct placement of what was seen in the original image but from a 180 degree turn around.

This is an invaluable process and why Krita is an essential part of my toolkit. The video above gives you a good insight into exactly how I do all of it.

For advanced use and other examples I will post more in the future to my Patreon, so join that to be kept informed.

Workflow: To download the QWEN-Krita ComfyUI workflow shown in this video right click here and download the json file. Drop it into Comfyui and it will load up. (This is just another way to approach using QWEN in ComfyUI, but borrows the method from ACLY plugin in Krita, hence the naming)

back to top of page

DETAILERS

To be kept up to date with latest workflows and information join my free patreon tier

There are older still valid workflows that can be found in Research 2025. Detailers differ from upscalers because they apply models to the video instead of just resizing them, though they generally detail and upscale at the same time.

LTX-2 DETAILER

NOTE: The LTX-2.3 version is now available as part of the Video Pipeline Workflows pack and can be downloaded from that section. - Mark, March 2026

Video: A video discussing this workflow can be found here on Youtube channel. If you join the Patreon Paid Tier there is a part 2 video where I discuss the settings further and how I use it. You do not need that to use this workflow, and there are notes in the workflow.

Date: 26th February 2026.

About: This will be my detailer/upscaler moving forward.

(CAVEAT: it wont work for dialogue scenes, it will strip out mouth movement. This will only work where you dont care about changes happening in the results. Check the USDU detailer if you want lower impact detailing.)

I input a small video of 480 x 277 (16:9) into the inbound video node and then run it through either the GGUF model or FP4 to get a result at 1080p (actual: 1920 x 1024).

After numerous tests this has proved to be the best I can achieve with my current hardware. I would use the HuMO detailer from AbleJones shared in the next section, but I cannot reach 1080p with it and 720p didnt beat this LTX-2 workflow.

Note that if you are lowVRAM this LTX-2 detailer will OOM on the VAE Decode so you need to set your comfyUI up to enable you to push through without it restarting. This usually just requires the models remain loaded (see the switch settings in the About section for how I do that).

This is also the first workflow that features the new LTX node layout required to address the coming breaking change that ComfyUI are planning with LTX GGUF loading workflows. Details on that are in the notes of the workflow. A slightly confusing response came from ComfyUI devs to my question on it here. I dont believe it will break GGUFs, only the nodes, but it remains to be seen when they finally implement it.

Workflow: To download the LTX-2 Detailer-Upscaler-V2V workflow right click here and download the json file. The workflow can be dropped into Comfyui.

NOTE: Credit goes to VeteranAI for this workflow as it is based on his original idea.

back to top of page

HuMO DETAILER

Date: 11th February 2026.

Updated 26th February 2026: I tested this workflow of AbleJones further and discovered it is very good, better than LTX-2 but I cannot get to 1920 x 1080 with it and below that size it doesnt do so well. I won't be using this only because my hardware can't use it at maximum value. If it could I would use this for detailing. Alternative is USDU detailer which will allow even LowVRAM to achieve a result and you could adapt it to use HuMO but I havent because it takes too long to finish with WAN.

About: Detailers are v2v workflows. For example, when you have a finished video out of LTX-2 but you need to add detail or fix minor issues, this remains one of the best ways to "polish" or fix blemishes, effectively giving it a second pass through an alternative model for a touch-up.

At this moment in time, I cannot get this workflow to go above 480p, but it has one very interesting feature I was not aware of with HuMO - character consistency. (Like Phantom or MAGREF models).

This workflow from AbleJones (a.k.a. Droz) has been cunningly designed by him to use the first frame of your inbound video to inform the rest of the video about your characters. Thus you can force your characters back into the video based on the first frame (if you use FFLF this will usually be high grade).

This is fantastic, but to really make use of it you need to get to 720p or more. Right now I cannot with this workflow. Why? Probably because it also features the equally fantastic ClownShark Sampling stuff, but unfortunately for me on a 3060 RTX ClownShark stuff weighs heavily on time it takes to complete. Above 480p, it OOMs. So I need to tweak this workflow to get it reaching 720p before its of value to me, but you might find it useful.

Workflow 1: To download the HuMO full detailer vrs 3 (AbleJones) workflow right click here and download the json file. The workflow can be dropped into Comfyui.

Workflow 2 (26th February 2026): This is the butchered test workflow I was using for uploading LTX videos in and detailing them. I would use this if I could get to 1920 x 1080 with it on my hardware but I can't so I settle for the LTX-2 detailer-upscaler. This is vrs 5.5 of AbleJone's HuMO detailer but I kept only what I needed. right click here and download the json file. The workflow can be dropped into Comfyui.

back to top of page

USDU DETAILER

NOTE: This is now available as part of the Video Pipeline Workflows pack and can be downloaded from that section. - Mark, March 2026

Ultimate SD Upscaler (USDU). This is a great detailer for polishing or slightly fixing v2v to 1080p. It can use WAN or LTX models. But please note it is a WIP and can be extremely quirky especially when you change small settings (denoise, steps, or tile sizes).

Video: See the video discussing this workflow here . If you join the Patreon Paid Tier there is a part 2 video where I discuss the settings further and how I use it. You do not need that to use this workflow, and there are notes in the workflow.

Date: 28th February 2026.

About: This works best with WAN but it takes a long time to complete. LTX works faster (15 mins vrs WAN 40 mins) but results can be quirky in LTX. It is a WIP.

You would use this detailer when you have a mostly working video and need it polished. Low denoise (0.1 to 0.35) allows less risk of changing things like mouth movement but the challenge is finding the balance. Different models also require different settings.

My intention making this was to fix faces at a distance or provide an answer to v2v where mouth movements need to be maintained. With WAN model it does well, but it takes too long. It also has coloration issues on 81 frame change due to WAN model limitations. This could be addressed, I havent had time to look into fixing its issues, nor will I. Why? Because there will be better solutions down the road that require less time to get right.

As of Feb 2026 this is about as good as it gets. I would recommend using WAN if you can in this workflow, or the LTX detailer if you want better results and dont care about mouth movements or changes in the shot.

This workflow fills the gap where WAN is needed in LowVRAM (I did 233 frames at 1080p in 40 mins), and where you can't risk too much getting changed, also it might work where dialogue is present (not effect mouth movement so much). How well it works, remains to be seen in use. I only just finished making the workflow and testing it.

Workflow: To download the USDU WAN/LTX workflow right click here and download the json file. The workflow can be dropped into Comfyui.

back to top of page

WAN DETAILER

Date: 11th February 2026.

About: Detailers are v2v workflows. For example, when you have a finished video but you need to add detail or fix minor issues, this remains one of the best ways to "polish" or fix blemishes, effectively giving it a second pass through a WAN model for a touch-up.

A good example was an Eastern Brown snake in LTX-2 ended up with a head more like the snake in the Jungle Book cartoons, despite providing a good First-Frame image of the sname in the FFLF workflow. Running the final LTX-2 video through WAN at low denoise setting with a simple prompt: high quality, photorealistic. An Australian eastern brown snake. then setting it to 0.78 denoise adjusted the snakes head to a more correct look and tidied up the entire video.

Lower denoise levels between 0.3 and 0.79 can polish or be used to drive stronger fixes without losing the content of the original video. From 0.8 denoise and above, you will end up with completely different results than you put in. Still those are often good, and driven by the prompt, if you want that.

The issue now with this method is WAN is 16fps, but I find 157 frames (24fps out of LTX-2 reduced to 16fps = 157 frames) will work at 480p for this approach without issues. But going for higher resolutions runs into problems on my 3060 RTX at 157 frames with WAN. (now resolved to 1080p with the LTX-2 Detailer).

You can use any model you like but t2v models are best generally. WAN or any type will work. I use WAN 2.2 t2v Low Noise model in this example workflow just because it always served me well. Using reduced GGUF models might allow for 720p, but I haven't tried. (In "Footprints In Eternity" I used a 1.3B WAN model for this task).

Workflows: To download the WAN detailer workflow right click here and download the json file. The workflow can be dropped into Comfyui.

back to top of page

EXTENDING VIDEOS

To be kept up to date with latest workflows and information join my free patreon tier

HOT TIPS FOR LOW VRAM: When using a workflow with previews in I recommend running it with the ComfyUI browser moved to an empty part of the screen. Tests on my setup have shown me that GPU "blap" is being caused on my lowVRAM by trying to watch video previews. This causes huge impact on speeds. I recommend process tracking your GPU constantly to know what it is doing.

LTX-2 EXTENDING DIALOGUE/LIPSYNC WITH REF IMAGE & AUDIO-IN

NOTE: My current video pipeline is available as part of the Video Pipeline Workflows pack and can be downloaded from that section. I am no longer using LTX-2 models (I use 2.3) but leave this here for people who are - Mark, March 2026

(A brief video for extending dialogue using this workflow is on Patreon (free tier), a longer workshop discussion is in the paid tier.)

Date: 17th February 2026

About: This workflow is i2v with audio driving the lipsync. It is based on the same original wf in LTX-2 EXTENDING FROM IMAGE OR VIDEO CLIP but this one seems to work better with dialogue and scenes without random changes, but I think its a bit slower than the other one. At this time it is unknown exactly why it has these differences, probably settings. (This wf comes from Purzbeats and Torny, the other Kijai one I got from discord channel some time back and adapted. Both are fundamentally similar).

This workflow worked best for extending my dialogue scene with two people and will extend a scene several times quite cleanly (up to x5, at x6 it misbehaved).

It is recommended you minimise the prompts to avoid it creating random scenes. Use only the first prompt for all extensions (I added more in the Kijai version but later disabled them), else it will carry the early prompts across and change scenes.

The way this model works to extend, it will to try to drive the people back to the first frame position each extension start. Fight that if you must, I don't bother.

Therefore with dialogue it helps to keep the camera still (dolly in post), and keep the prompt to a minimum. There is also some decision-making to be done about the Loras that have an effect on human interaction, I have provided notes in the workflow beside the loras about that and will keep them updated with new results.

An example prompt for extending dialogue: a man and woman talking to each other. The camera is static, does not move.

Workflow : To download the updated dialogue/extension i2v workflow right click here and download the json file. You can drop it into comfyui. Your ComfyUI will need to be up to date (Kijai custom nodes as of 13th January) for the additional nodes to work.

back to top of page

LTX-2 EXTENDING FROM IMAGE OR VIDEO CLIP

NOTE: My current video pipeline is available as part of the Video Pipeline Workflows pack and can be downloaded from that section. I am no longer using LTX-2 models (I use 2.3) but leave this here for people who are - Mark, March 2026

Includes two workflows using different approaches to extend a video. This one didnt work so well as the other with dialogue as the scene changes, but might be useful in other contexts.

LTX-2 Extending Videos - Two Approaches

Date (Original): 24th January 2026.

Update 1: 11th February 2026.

The Kijai extension workflow originally used in the video above for the Mel Gibson Braveheart scene has been updated to include some extra features - mostly memory improvement nodes. The RuneXX workflow is still provided but not recommended as it has issues with frame counts 8+1. You can work around it but I no longer recommend that workflow. Though it is the only LTX-2 v2v I offer at this time so can still be downloaded below. I will test others in the future, time permitting.

About (Original video and workflows): Two workflows are provided. (You will need to update ComfyUI to 23rd January 2026 (AEST) and Kijai nodes as well to make use of all the nodes featured).

In this video I discuss two different approaches to extending videos and provide both workflows in the links below.

  1. Using an existing video clip to drive a v2v output. This takes the original clip and blends it in with whatever you prompt for action and dialogue. The result is an extended video. (MBEDIT-LTX-2_V2V_Extend_RuneXX_Vrs5.json)

  2. Using a masked base image but driving "infinite" extension with an audio file for the lipsync dialogue. In this case I use a 28 second long audio file. (MBEDIT - KJ-wip-ltx2_extension_testing_vrs5.json)

The result can now get me to 720p on my lowVRAM (3060) GPU at infinite length (28 seconds took 40 minutes) thanks to additional nodes that I show in the i2v (Kijai based) workflow, these are recent additions to ComfyUI from Kijai, as well as a VAE memory improvement from another dev Rattus (available in update to ComfyUI after 23rd Jan 2026 AEST).

The power of the LTX model is only just starting to be understood and is rapidly evolving with all the dev attention. So this is a work in progress for both workflows, and both could be adapted to take alternative approaches as well.

Workflows (original): (NOTE: some people have run into issues with the RuneXX workflow frame count erroring with 8+1 errors for latent frames. As such I no longer recommend using this one (v2v). I leave this available for your experimentation only.) To download the two (original) workflows shown in this video right click here and download the zip file. The workflows can be found in the json files and you can drop them into Comfyui and they will load up.

Updated Workflow (11th Feb 2029): To download the updated i2v workflow right click here and download the json file. You can drop it into comfyui. Your ComfyUI will need to be up to date (Kijai custom nodes as of 13th January) for the additional nodes to work.

back to top of page

FFLF (First Frame, Last Frame)

To be kept up to date with latest workflows and information join my free patreon tier

NOTE: My current video pipeline is available as part of the Video Pipeline Workflows pack and can be downloaded from that section. I am no longer using LTX-2 models (I use 2.3) but leave this here for people who are - Mark, March 2026

LTX-2 FFLF

NOTE: My current video pipeline is available as part of the Video Pipeline Workflows pack and can be downloaded from that section. I am no longer using LTX-2 models (I use 2.3) but leave this here for people who are - Mark, March 2026

LTX-2 FFLF (First Frame, Last Frame)

Date: 21st January 2026.

Updated Workflow: 11th February 2026.

New nodes and loras have been added in this updated workflow and offer improvements to the original. Nothing else has been changed so you can use either the orginals or this one.

About: I really like this workflow and the look of LTX-2. The FFLF works well in this one and I tested a lot before finding it. It's also easy and quick to get to 720p and that is where the "Blancmange" effect starts to disappear, especially in this workflow.

Since making the video I have added the NAG in which now provides negative prompting (neg prompting wont work with cfg 1 without it, and cfg 1 is needed for the distilled models which I use).

More notes in the workflow will help you know how to use it.

I havent included Lipsync in this one because there are problems with lipsync audio-file-in approach but that will be the subject of another workflow and video. The issue has a solution but didnt work well in this workflow. But generally my dialogue shots wont be FFLF so its not something I bothered to address here. One workflow for each task suits me fine.

This workflow comes from Phr00t, I didnt design it. He has done a great job. Link to his github is in the workflow.

Workflows: To download the original workflow shown in this video right click here and download the png file. The workflow is in the metadata of the png file and you can drop it into Comfyui and it will load up.

Updated Workflow: this updated workflow has the latest memory improvement nodes, loras, and tweaks. If it doesnt work for you, the original is above and will work just fine. This has improved features worth using though. right click here to download the updated LTX-2 FFLF workflow and drop it into Comfyui. You will need Kijai custom nodes to be updated (after 13th Jan 2026) for the additional nodes to work.

back to top of page

INSTALL COMFYUI PORTABLE

To be kept up to date with latest workflows and information join my free patreon tier

Fresh Install Of Comfyui Portable - How I Do It

This is specifically for Windows 10 version and my hardware setup. You will need to do your own due dilligence to ascertain what will work best for your hardware and operating system.

Date: 8th March 2026.

About: This is 50 minute video talking through the process I went through. I am not an expert. This is just my approach taken when I installed a fresh comfyUI portable install on my Windows 10 machine.

CAVEAT: this video is a guide detailing my experience and sharing how I installed ComfyUI portable on my setup on 5th March 2026. Your hardware will require different things. I am NOT available for tech support. I am NOT responsible if you destroy your machine blindly following what I did here. Do your own research, but this will help guide you to the places you need to install it.

Sage Attention and Triton used to be horrible to install but if you use the links below and your hardware meets the criteria they are now a lot easier.

I am now upgraded to a fresh install of comfyui-portable running on Windows 10 with python 3.13, pytorch 2.10, CUDA 13.0, Triton 3.6 installed for a 3060 RTX GPU using NVIDIA studio driver 591.74 and as of March 2026 some of that really is the baseline requirement for LTX-2.3 to work best, which is my main focus at this time.

Links mentioned in the video:

ComfyUI install instructions: https://docs.comfy.org/installation/comfyui_portable_windows#download-comfyui-portable

Triton wheels and install: https://github.com/woct0rdho/triton-windows
Sage Attention wheels and install: https://github.com/woct0rdho/SageAttention


Follow me on Patreon to be kept up to date. And if what I do helps you, buy me a coffee by paying for a months membership. It helps me to keep all this going.

back to top of page

LIPSYNC

To be kept up to date with latest workflows and information join my free patreon tier

LTX-2 LIPSYNC (Using Audio File)

NOTE: My current video pipeline is available as part of the Video Pipeline Workflows pack and can be downloaded from that section. I am no longer using LTX-2 models (I use 2.3) but leave this here for people who are. - Mark, March 2026

LTX-2 Lipsync Audio-In (Includes fixes for frozen-frame output)

Date: 22nd January 2026.

About: The most important thing is that this workflow fixes the "frozen frame" issue. It also introduces a fix for the fiddly solutions required when using the static camera lora (which I left in as some might find it useful).

The best solution is adding in the distilled lora and setting it to -0.3, once I did that I no longer had to keep tweaking multiple settings. The distilled lora is 7.5GB but weirdly still runs fine even on my 3060 RTX 12 GB Vram with 32 GB system ram. I havent fully tested this workflow beyond basically running it at various resolutions and no longer having all the problems show up. Consider it a wip that might be a final copy. If I find any extra things I will add them in and update the workflow below, and include a date of upload.

Notes in the workflow should get you started. As always I dont provide exact location of models as I assume you will want different ones based on your hardware. Most models come from three or four sources: Kijai, Quantstack, City96, and Unsloth all found on Hugging Face. (That is the GGUF's for us LowVRAM peasants, if you are using the big boi GPUs, you will likely have other sources).

The base workflow came from a guy called Abyss on discord, I have no idea where he got it from and have adapted it to suit my purposes.

Workflow: To download the workflow shown in this video right click here and download the png file. The workflow is in the metadata of the png file and you can drop it into Comfyui and it will load up.

back to top of page

UPSCALERS (1080p)

To be kept up to date with latest workflows and information join my free patreon tier

This is different to Detailers in that it is specifically for upscaling

FLASH-VSR

Date: 11th February 2026

About: How did this thing bypass my radar? I'll tell you, because I installed a version that was not this good. I am impressed by this upscaler.

My 3060 can push LTX-2 output video at 720p, 24fps, 241 frames (10 seconds long) to 1080p in just 10 minutes with great clarity. I have to experiment with it further to find the sweet spot between over clarifying and blending in what I need.

This version is from "naxci" - https://github.com/naxci1/ComfyUI-FlashVSR_Stable (I have used others before and didnt get the results as good as this)

Make sure you download the models that are version 1.1. They are named the same, come from the same Hugging Face account, but annoyingly I downloaded 6GB of original models before realising the new ones were out. Follow the instructions in the naxci link to get it up and running.

Workflow: To download the workflow right click here and download the json file. The workflow you can drop into Comfyui and it will load up. You will need to install the relevant custom node first.

back to top of page

USEFUL SOFTWARE

100% open source software (actually Davinci is not OSS, but is free with licensing caveats)

back to top

VIBEVOICE

To be kept up to date with latest workflows and information join my free patreon tier

VIBE-VOICE & AUDIO SEPERATION

NOTE: As of 18th March 2026 VibeVoice the ComfyUI manager version of this will install transformers > vrs 5 and it will not work. You need to downgrade the Transformers to a suitable version then it will work. I have a github note for it here explaining the fix until the dev attends to the issue.

Both Vibe-Voice and Audio Seperation feature in the same workflow.

I did a recent TTS shoot out in the Patreon Free Tier between QWEN 3 TTS, VibeVoice TTS, Chatterbox TTS. I expected Chatterbox to win, it didn't. I remain a fan of Vibevoice at this time.

Date: 11th February 2026

Update (17th February 2026): I have now included Multi-speaker node and also FL Audio Seperate nodes (from Fill) into the workflow. Disable or delete the sections if you dont wish to install the Fill custom node, but these are very useful audio seperation tools I had no idea existed yet were installed in my setup.

The VibeVoice multi-user works really well for dialogue between men and women in LTX without any prompting. LTX manages to know who is speaking if the tonality is different enough and applies it correctly.

About: This is not a unique workflow but is the VibeVoice version from Enemy-X Net and has the MelbandRoFormer nodes added in (from Kijai custom nodes found in the manager) for Audio Seperation (voice vrs instruments).

I use the MelbandRoFormer for two things:

  1. Cleaning up the background noises VibeVoice tends to make

  2. Seperation music tracks into vocals and instrumental for analysis or "narration" over the top to remove distracting vocals. (Do not use copyrighted songs for YouTube unless you know what you are doing. I use this for testing and research, or to analyse music tracks and vocal production methods)

I hadn't realised how good MelbandRoFormer was until I tried it in this setting. Now I use it in all my vibevoice TTS runs as it also normalises the output levels.

Workflow: Its best to install it from the Enemyx-net site linked above and follow their instructions for use of models and so on. But to download the workflow (updated 17th February) right click here and download the json file. The workflow you can drop into Comfyui and it will load up.

back to top of page

VIDEO PIPELINE WORKFLOWS

Due to the constant time-consuming need to maintain this website with updates, I will henceforward share new and updated workflows only via the Patreon (free tier). They will be accessible to everyone, even non-members. - Mark, 16th April 2026

LTX-2.3 VIDEO PIPELINE WORKFLOW

Date: 18th March 2026

About: These are the workflows I use in the video-pipeline for LTX 2.3 to get final results. They are the same as I was using for LTX-2 but updated to cater to minor changes and the new LTX2.3 models (I use LTX 2.3 dev Q5_KM model with kijais adatped distill lora set to 0.6)

The pipeline:

Workflows 1a and 1b I use to make the initial structure. I work at 480 x 201 (2.39:1 aspect ratio) because its faster than 16:9 ratio, and also it gets me the structure I want quickly running again if not, before I upscale it to a final video. Workflow 2 is where I upscale the 480 x 201 twice through upscalers to get me to 1920 on the longer edge. If the result is still lacking something or has minor quality issues but is otherwise good, then I run it through USDU WAN driven workflow 3, but it takes 25 minutes so I only use it if I need it. But even on LowVRAM that will finish a 1080p, 241 frame video if you adjust the tiling to a 3:2 ratio and use 1 step. It is a great final polish trick making use of WAN low denoise to avoid collapsing the consistency, but is slow.

Workflows: The json files for the four workflows are in a zip file. right click here and download the zip file. Unpack it and the workflows you can drop into Comfyui and they will load up. You will need to select your own models and file path locations.

File names & Versions in zip file:

1a. MBEDIT-i2v_kj_LTX23_extend_audio-in_devQ5_vrs3.json
1b. MBEDIT-LTX23_FFLF_NAG-New_DevQ5KM_vrs3.json
2. MBEDIT-v2v_LTX23_Upscaler_DevQ5KM-Audio-In_1080p_vrs7.json
3. MBEDIT-USDU_LTX-2_Or_WAN22_Detailer_torch210_fix_vrs1.json

back to top of page

We use cookies to improve your experience on our website. By continuing to browse, you consent to our use of cookies. Learn more