RuneXX

/

LTX-2.3-2.5-Workflows

like 843

I also have this issue, the images from the latter scenes are used in scene two, for example.

Will check. I did see at some comfy builds it seemed to "cache" the image loaders, and got confused.

Try update comfyui so we are at same version.
I tried reproduce it my end, but everything ran smootly .

( i was able to reproduce it earlier, and was on my list of things to try other image loader, since it seemed like ComfyUI cached or did something strange with multiple image laoders, but when i am trying now, i cant reproduce it, everything seems ok again)

Try update comfyui so we are at same version.
I tried reproduce it my end, but everything ran smootly .

( i was able to reproduce it earlier, and was on my list of things to try other image loader, since it seemed like ComfyUI cached or did something strange with multiple image laoders, but when i am trying now, i cant reproduce it, everything seems ok again)

So I have updated my ComfyUI to the nightly update as i usually do. And the first scene is always merging with the fourth scene and so on. The images are merging together in the scenes.

Ok, will see if i can reproduce. I did have this happen once, that comfyUI got "too lazy" to load all the images.
But will try a better image loader than the stock one, should hopefully do the trick ;-)

(comfy gone a bit too extreme on some caching perhaps)

Try this one:
https://huggingface.co/RuneXX/LTX-2.3-Workflows/tree/main/Music-Video-Creator/Image-Test

Basically what it does, is forcing comfy to load all the images earlier by making a little image collage.
The theory being that the other images are loaded so late in the workflow run, that comfy gets "lazy" and just picks one from earlier loaders (some cache issue)

(if that doesnt fix it I will use other image loader instead.. actually maybe i can make one already quickly to test just in case)

Gonna load it up and test it right now :3

Gonna load it up and test it right now :3

Fingers crossed ;-) and if you downloaded before yet another little fix, try download again (see the date on the file upload)

Actually curious if you had issues with the "low ram" (that saves each segment as it own file and merge at end), or the "full" workflow .. that does all in one go.
Reason for asking is that the "full" had a connector between nodes, to connect to multiple (very common to use), but since this connector was connected to the subgraph input, Comfy saw this as a as a "general could be anything input" instead of a dedicated image input. Maybe that was it. But the test has a fix for that as well (plus the force load images earlier)

Gonna load it up and test it right now :3

Fingers crossed ;-) and if you downloaded before yet another little fix, try download again (see the date on the file upload)

Actually curious if you had issues with the "low ram" (that saves each segment as it own file and merge at end), or the "full" workflow .. that does all in one go.
Reason for asking is that the "full" had a connector between nodes, to connect to multiple (very common to use), but since this connector was connected to the subgraph input, Comfy saw this as a as a "general could be anything input" instead of a dedicated image input. Maybe that was it. But the test has a fix for that as well (plus the force load images earlier)

So i was using the full one. And right now i just had downloaded the one before the tiny fix you just did and it still did the error with the merging of images

Just tested the new another fix one, its still putting scene 4 into scene 1

I think it might be the img compression, cause on your other workflows you have the img compression set 1, and this one is 18. When i change it to 1 on the workflow, it doesnt mix images on this new fix

So i was using the full one. And right now i just had downloaded the one before the tiny fix you just did and it still did the error with the merging of images

The full test one you are using now, at segment 2, 3 etc.. The subgraph node, where ref_image connects. Does it have a blue dot? or green?

Was hoping the green dot was the cause, but if you have blue, I'll try a different image loader or different way. (green being "generic" input)
(i dont think its the image compression, its just there to "fool" ltx to think the image is a video frame, instead of a still image.. as per LTX own way of doing it. To prevent still image video output)

Its a blue ref_image dot

And just to make sure, update KJNodes to the very latest
(but i think the wf would perhaps error if you are not. It uses set and get nodes from parent into subgroup, something KJnodes added support for a few weeks ago)

Blue it is. crap.. ;-) will try some other way.

Looks like you might be on nodes 2.0.. but i doubt that matters

(the extreme fix would be to stitch all images together at the very start, and then crop out the needed part at every segment. That will definitely work, no way that could fail. Will do one for that too as a last option. I think the problem might be that comfy gone a bit overboard with caching, and assuming noone loads more than 2-3 images. But i'll digg into it and figure it out)

Everything has been freshly updated. I use a rental gpu via vast.ai to do this so have to clean install nodes everytime i boot up a gpu

The strange part is that it works for me, and others too. Some guy made a music video and posted a few hours ago:
https://huggingface.co/RuneXX/LTX-2.3-Workflows/discussions/92#69f3745e284357cd15723f09

But i have ran into the issue you are having, once in the past. So i know there is something... . I'll figure it out ;-)

Im also going to test out the low ram version and see if that works or not. It might just be the gpu im using even though its high in vram just might not like the full version lol

So far, the low ram version seems to work without any issues o.o

So far, the low ram version seems to work without any issues o.o

really? wow.. the plot thickens, but that good help at figuring it out. I'll take a close look at what could differ between the two workflows.

Btw, i will also upload a different method for making music videos also, with prompt relay (basically its 1 video, no multi video parts, the segments are done in the prompt itself).
A viable alternative ;-)

But if the low ram works, i'll look for the needle in the noodle haystack, see what it can be for the "regular" version that could make images not always load correctly ;-)

The similar problem was for me: https://huggingface.co/RuneXX/LTX-2.3-Workflows/discussions/92#69eb90238d2116116d9ddad6
Looking at the wf names (LTX-2.3_-I2V_T2V_Music-Video-Creator_multi-scene_custom_audio.json, LTX-2.3-I2V_T2V_Music-Video-Creator_multi-scene_custom_audio_Low_RAM.json) I thought that low RAM memory was not my case. So first I used "LTX-2.3-I2V_T2V_Music-Video-Creator_multi-scene_custom_audio.json" but this gaves the problems. The "LTX-2.3-_I2V_T2V_Music-Video-Creator_multi-scene_custom_audio_Low_RAM.json" maintaining functionality provides the temporary scenes files during complilation. I thinks the wf naming slightly misleads.)))

So far, the low ram version seems to work without any issues o.o

really? wow.. the plot thickens, but that good help at figuring it out. I'll take a close look at what could differ between the two workflows.

Btw, i will also upload a different method for making music videos also, with prompt relay (basically its 1 video, no multi video parts, the segments are done in the prompt itself).
A viable alternative ;-)

But if the low ram works, i'll look for the needle in the noodle haystack, see what it can be for the "regular" version that could make images not always load correctly ;-)

Yeah it is quite weird about this O.o i examined between both of them and couldnt find anything myself out of place. So good luck :D and im glad i could help out some ^^

So far, the low ram version seems to work without any issues o.o

really? wow.. the plot thickens, but that good help at figuring it out. I'll take a close look at what could differ between the two workflows.

Btw, i will also upload a different method for making music videos also, with prompt relay (basically its 1 video, no multi video parts, the segments are done in the prompt itself).
A viable alternative ;-)

But if the low ram works, i'll look for the needle in the noodle haystack, see what it can be for the "regular" version that could make images not always load correctly ;-)

Low ram version might be way to go for longer videos with higher resolutions, I will try to extend number of segments today in low res wf. The first version of WF u made (non low vram) for me limit is 115 seconds at 720p and it crashes at combining all frames as soon as that length 115 sec is crossed (system with 16gb vram and 48 gb ram) while that wf does 20-30 sec longer music videos easily at 480p but then u really cannot upscale it well to 1080p which is my goal. I mean one can stitch segments in D. Resolve or Shutter encoder what matters to me is that getting all segments made is automated, stitching them together can be done manually - that's quick thing to do.

Just a quick example, using the Prompt Relay with custom audio workflow instead for same ish goal... music video.
An alternative wf, while i try figure out the "music video" one failing to load images in correct order for some.

If you wanna try : https://huggingface.co/RuneXX/LTX-2.3-Workflows/tree/main/Movie-Maker/Prompt-Relay-Custom-Audio

(the video was just a quick example not at full res, but if one is patient to wait a bit, for higher res, or even more steps, the result can be even nice ;-)

this was low rest test just to prove it works fine for long videos, I need to redo images to get more char consistency but it terms of style it delivered what I wanted so only new images and then 720p vid. generation is next task..

yes the wf can be used for long videos, easily to 3 minutes with 8 segments but it required "ploughing" for a half a day through wf to make it possible, default it cannot do it properly, not sure why forcing images is needed in this WF it works without it just fine first thing that was removed, then the final merge in default wf, it does not work properly, VHS nodes create audio version of mp4 and mute mp4 without audio track in temporary folder so all get merged and even then I kept getting message the folder does not exist although it did exist. It is something I noticed when first testing the low vram wf a week ago. So that section was gone. I just needed segments. I increased number of segments to 7 , I needed that number for video I already had images for and music and prompts prepared. Flaws in the default wf one that I instantly spotted while modding was that get window sec 3 is not connected to simple KJ calculator. It is present in all wf as such. The other thing is for long videos in low vram wf formula a + b + c + d + e + 2 in that simple KJ calculator has to be replaced with a+b+c+d+e.. without adding compensating number 2 meanwhile formula in simple KJ calculator in Graphs of each segments has to be changed from ((round((a * b -1) / 8)) * 8) + 1 to >>>> ((ceil((a * b -1) / 8)) * 8) + 1 , it just works better after u cross 2 minutes mark of length, for example this track I used is 132 seconds long with original rounding formula resulting video with 7 segments was 131 seconds long, audio length with original formula was 134 seconds while original audio track is 132 seconds long the result was was that there was unnatural jitter jump between segments without breaking lipsync, with formula I used resulting video is 132 seconds long and after merging segments saved without audio track with shutter encoder and then adding original audio track to that video I got result posted here it looks smoother and u can easily add cross dissolve in D.Resolve to improve it further so pipeline for this video was (the formula u could say pads the end of segment with 5-8 frames), so back to pipeline: modded wf - > merge with shutter encoder (segment videos without audio) > Add original Audio track with Shutter encoder to merged video > Add cross dissolve between segments in D. Resolve > Topaz Video AI to add film grain and refine faces just a bit as it was low res video. Further testing is needed to see if changes will produce consistent results for long videos.

That video looks really nice ;-)

Yes the VHS node was made prior to LTX and video models having audio. So it saves a mute video, and a png (workflow embedded), as well as the proper video with audio.
I turned that off long ago. You can turn off save meta data at the node, and in the settings for that node, turn off save intermediate file and png ... something like that if i recall correctly.
(I'll swap to a different way of saving the video though, so that users that have not turned this off, it will work. Using the KJNodes save video instead)

In process of optimizing this music video workflow since some had issues. Will look over what you said as well ;-)

the final video, should have kept it all medium shots and close ups to prevent face melting in distance but it was already cooking - so no interrupting ...next task is to go for 3 minutes long video and 8 segments.

That looks amazing ;-) nice nice
That looks like a proper music video ... well done ;-)

(some bug with huggingface, not getting notification of new posts_didnt see this one before now)

Oh God this video is AMAZING!
Been starting a Ai Band featuring me and my 5 years old daughter and planning to make music video using your workflows with original lyrics. This is exciting!
I have been editing digital videos for 25 years and this is just the pinnacle of Tools to create music videos.
Not to hijack this thread, but id like to share this simple workflow to generate song variations quickly with 2 styles input etc.
https://huggingface.co/WanApp/UltimateMusicMaker

Thank you for all your work RuneXX, if you come by Montreal some days, poke me, I'd like to pay you a coffee or a good restaurant!

That looks amazing ;-) nice nice
That looks like a proper music video ... well done ;-)

(some bug with huggingface, not getting notification of new posts_didnt see this one before now)

btw reasoning loras break and overrule prompt in Prompt Relay Encode / Timeline/Lipsync wfs ( Looks like node issue but isn’t) and act as antagonists in anything that isn't humancentric so to say, u can say goodbye to special effects, movement based on prompt as the reasoning lora will derive movement from pose on image and override model from doing what is prompted (example u have two people looking over shoulder at the camera in the frame and want them to pivot and turn their bodies toward camera and walk towards camera and leave the frame that way, without reasoning lora it will happen as it was prompted, add reasoning lora they will go towards the door and leave frame that way by disappearing through the door in front of them because the reasoning lora will prioritize pose and derive movement from that so faces looking at camer okay..but bodies directed towards the door it makes them walk to the door..it does not matter what str u use it will be an antagonist and in music video making wfs with lipsync it will try to sing on instrument tones, when I made this test video last night with 8 segments dystopian grunge metal with Severance tv show visual style, just a quick test of style concept, with LICON reasoning lora at .5 str singer was trying to sing guitar solos 👹 and visual effects prompted per segments were not showing up, second run without lora all went perfect. I had to redo all videos I made with reasoning loras, oh boy what a difference that made visually!

That looks great ;-)
Well done.... I often sing solos myself haha

I guess no reasoning lora then ... hehe.. Only tried the reasoning lora a few times, but then it seemed to help (it was not a music video though). Helped with logic such as opening doors and entering rooms etc.
But often less is more, keeping it barebone less things that might alter the output ;-) And could be that the reasoning lora in Prompt Relay combo is a bad idea true

That looks great ;-)
Well done.... I often sing solos myself haha

I guess no reasoning lora then ... hehe.. Only tried the reasoning lora a few times, but then it seemed to help (it was not a music video though). Helped with logic such as opening doors and entering rooms etc.
But often less is more, keeping it barebone less things that might alter the output ;-) And could be that the reasoning lora in Prompt Relay combo is a bad idea true

yeah it helps in complex movement scenarios if you cannot pull it off normally with prompting like opening door, window using spoon, fork etc, but it works ok in i2v single image and single prompt, no relay encode, no timeline, plain vanilla template wf, it was designed before all these new nodes we got recently and it just does not get along with them or acts as an antagonist vector, suppressing model creativity and putting priority on humans and their movement at the cost of scene coherence and prompt adherence, first victim are special effects, there was no rotation scene at 2:14 for example and it would block creation of all prompted dust particles but funniest thing was trying to lipsync guitar solos as if it was a scream. Maybe we'll get new versions that will work better...time will tell.

Watched your latest video again.. its super ;-)
The camera movements make it diversify .. really nice ;-)

havent tried a rock video myself yet, but that video makes me wanna give it a go.

(just a low res 720p example video)

Updated "low ram" save video per segment workflow :

• More accurate frames to audio : now it uses the actual calculated frames to set audio duration instead of the window seconds input. (since it follows actual frames, it shouldn't matter if its round or ceil)
• Better fps condition inside each group
• Better save intermediate files using native "Save Video" node (and not using VHS combine that can end up saving extra silent video and png for workflow meta data)
• Changed slash direction for directory path to match that of the node (should hopefully fix some OS not finding the path for the full video)
• "Window seconds" input removed, to make extending groups more easily (and the trim audio input node at the very start is set to 2 minutes, but can set it higher or lower to match the workflow duration)
  (the trim audio node at the very start is also useful to continue the song with a 2nd, 3rd etc generation, by using the audio start from to match where the previous video ended)

Should hopefully work better ;-)

If it breaks at the very end where it combines all the saved intermediate video files, I'll upload a manual combine video "helper" workflow so one can easily combine the intermediate video files manually (the intermedia video files are saved even if the end auto combine node should fail, in a temporary folder).
This helper workflow also have optional frame interpolation and upscale video part.

Updated regular "music video" workflow
(this variant of the workflow does not save intermediate files, but rather generates it all in one go...)

• Removed previous batch image nodes with overlap and forward (these are more typical for overlapping frames, and this workflow does not have that)
• Each group's output frames stored in set variables, and merged at end with image batch multi for the full video (instead of accumulating frames and forward to next groups). Might be a little less RAM heavy.
• ... plus all the small things from above (more accurate audio duration, better fps conditioning etc)<
• A bit simpler alternative than the "save each segment" wf.

(not really sure both workflows are needed, since the save intermediate files wf might be the better option for many/most. But since it used to be 2 variants.. They are quite similar though)

Plus made an excuse for a little rock video when testing ;-) 🤘🎸🎶

This is final fine tuned video ready for YouTube with more dynamic audio track of the same song, once the reasoning lora was ditched it was easy to further enhance prompting. This was with old wf I will be testing the new WFs during the next few days.

Nice nice ;-) looks amazing
Ready for some YT for sure ;-)

Catchy video and catchy song

Nice nice ;-) looks amazing
Ready for some YT for sure ;-)

Catchy video and catchy song

Would T2V version of this WF be possible or one with master switch from T2V to I2V? Vrgamedevgirl has nice T2V style loras and T2V test music videos she posted during last few days on Banodoco look stunning . U can easily make music video set in a fantasy world just by using her loras (all available on Civitai).

Think it already is quite ready for T2V. Will give it a test

I'm also struggling with the T2V - setting image strength to 0 doesn't seem to effect it. Also tried bypassing image nodes. I got the one from RareTutor to work for a single vid.

Yes might just need a little node change. Will take a look at T2V

Music Video - Prompt Relay combo

Uploaded a workflow variant of the "music video" workflows with Prompt Relay per group.
Makes it easier to variate a little bit, even within each group.

Its a lot of prompting though ;-)

(the video above is far from as much variation you can do, it was more a test run to check if all was working;-)_)

Music Video - Prompt Relay combo

Uploaded a workflow variant of the "music video" workflows with Prompt Relay per group.
Makes it easier to variate a little bit, even within each group.

Its a lot of prompting though ;-)

(the video above is far from as much variation you can do, it was more a test run to check if all was working;-)_)

oh boy, now I have to try this with licon reasoning lora just for the sake of science and giggles. 😄

oh boy, now I have to try this with licon reasoning lora just for the sake of science and giggles. 😄

haha 😄 might get all sorts of weird things, could be some lucky accidents haha

Was actually thinking trying some loras myself, the transition lora, and the music react lora. See what kind of trippy results i could get ;-) (and suitable for some sort of music, like trance etc)
In theory .. might just end up a mess ;-) haha

https://huggingface.co/joyfox/LTX-2.3-Transition-LORA
https://huggingface.co/100percentrobot/LTX-2.3-Audio-Reactive-LORA

Music Video - Prompt Relay combo

Uploaded a workflow variant of the "music video" workflows with Prompt Relay per group.
Makes it easier to variate a little bit, even within each group.

Its a lot of prompting though ;-)

(the video above is far from as much variation you can do, it was more a test run to check if all was working;-)_)

Aside from everything, I loved the song and the music <3

Aside from everything, I loved the song and the music <3

Try out AceStep XL ;-) its "Suno in comfy"
The songs I made were with Ace Step XL Turbo .. you can get even better quality with Ace Step XL STF (50 steps, cfg 4-7 ish, but still pretty fast)
Its good fun.. maybe not quite up there with Suno, but pretty good ;-)

Aside from everything, I loved the song and the music <3

Try out AceStep XL ;-) its "Suno in comfy"
The songs I made were with Ace Step XL Turbo .. you can get even better quality with Ace Step XL STF (50 steps, cfg 4-7 ish, but still pretty fast)
Its good fun.. maybe not quite up there with Suno, but pretty good ;-)

Just a quick question cuz I never tried it. it added the lyrics too?

Ace Step XL just generates the music.
So you have to provide 2 prompts.

• describe the music you want (style, instruments, genre etc)
• lyrics (same style as suno .. [chorus]... [verse]... etc)

For the lyrics, I just used Qwen Chat, but can use chatgpt, gemini etc as well. Just say a few keyword what you want the lyrics to be about etc. And the LLM does the rest ;-)
And now that Gemma 4 is supported at in ComfyUI, i might add that to the Ace Step workflow, it should be pretty good at generating both lyrics and the music style prompt for you

Would T2V version of this WF be possible or one with master switch from T2V to I2V?

Updated for T2V with a toggle between I2V and T2V ;-)

first video made with T2V, no post processing or retakes, first run where VRGameDevGirl's Painterly Lora was successfully used from start to finish.

Wow... that looks great ;-) really nice music video
A bit of a "fairytale" and "magic" feel to it... and nice story to it .. well done ;-)

That painterly lora i gotta try out, interesting output ;-)

Wow... that looks great ;-) really nice music video
A bit of a "fairytale" and "magic" feel to it... and nice story to it .. well done ;-)

That painterly lora i gotta try out, interesting output ;-)

this was just a test there is much better video being generated atm, Qwen 3.6 was used for that video it wrote wordy prompts for segments that LTX 2.3 just could not digest even when the system engineer prompt was adjusted and constrains defined, I plugged in Gemma 4 and I get what my vision was with 2 x shorter prompts and way more impressive visual effects. This Painterly lora is nuts, it can do crazy trippy stuff.

Gemma4's vision (prompt) of scene at 1:30 (low res test run)

Nice nice ;-) that looks more "layered" adding a bit of depth, really nice

Gemma 4 is also natively supported in Comfy now, so might be interesting to add in a few workflows where you need great prompts

Nice nice ;-) that looks more "layered" adding a bit of depth, really nice

Gemma 4 is also natively supported in Comfy now, so might be interesting to add in a few workflows where you need great prompts

Yes Gemma 4 just works nicer with ltx2.3 not that Qwen 3.6 does not have its place but it has different approach and does a few things its own way that might be exploited for some ideas.

Much better storytelling and adherence to my vision by Gemma 4 - this is now around 80% of where I want it to be some tweaks to prompts and a few retakes I will get it to cross line ..it has too many unicorns for my taste, need to replace them with spirits...I allowed 2 scenes without lipsync with wider shots and Gemma did decent job with it, some scenes have a few glitches, seed reroll needed.

Looks great ;-) nice little story video and catchy visuals and song ;-)

I gotta try out T2V mode, I rarely use that mode ;-)

Looks great ;-) nice little story video and catchy visuals and song ;-)

I gotta try out T2V mode, I rarely use that mode ;-)

She has just made it with her new WF that uses Gemma 4 ...this is with Painterly lora and pretty much proves how versatile it is and flexible but originally intended as Goth style inspired lora:
https://discord.com/channels/1076117621407223829/1458032975982755861/1501278750371741926

The link didnt work for me, probably you have to be "member" of her server/group to see it

The link didnt work for me, probably you have to be "member" of her server/group to see it

This is on Banodoco ltx gen community on discord, yes u have to join to see but it is public group also u have LTX2 resources section where all new stuff and experimental stuff can be found.

Will join the group.. if i can find it ;-)

She has just made it with her new WF that uses Gemma 4 ...this is with Painterly lora and pretty much proves how versatile it is and flexible but originally intended as Goth style inspired lora

Ah yes that looks nice ;-) Joined to watch it. Downloaded that lora (and a few others she had).
Will play around with it ;-)

I think u should add a note under trimming section of the first segment, useful for new users and beginners, that first segment can be used for retakes of itself and other segments, for example segment 4 u don't like it but like other segments u just place starting time of audio track of segment 4 in trim node and put length of video of that segment 4 in the first segment of the wf and copy prompt of the 4th segment to prompt window of the first segment and run VHS node of the first segment and u get retake done and can adjust prompt or re-roll seed. I am detailed and redundant in description for the sake of beginners so they don't get confused.

Yes that a good way of redoing a segment that didnt end up quite as you wanted.
Will add some tips about that ;-)

with Painterly lora. Low res test to see if It could do trippy stuff. 😁

hehe ;-) thats wild haha ... entertaining for sure ;-)

Monkeys from outerspace.. haha

hehe ;-) thats wild haha ... entertaining for sure ;-)

Monkeys from outerspace.. haha

inspired by this with pretty much the same lyrics https://www.youtube.com/watch?v=iHvrwUUvLZk

with Vrgamedevgirl's T2V Wild West Lora, I had so much fun making this and doing retakes.

Oh wow, thats insanely good ;-)
extremely catchy and the music video is superb ;-)

And great humor ;-) well done

Was trying out some AceStep XL loras, so ended up as a little video ... may the force be with you ;-) haha...
(low res, was just testing some stuff)

Updated regular "music video" workflow
(this variant of the workflow does not save intermediate files, but rather generates it all in one go...)

• Removed previous batch image nodes with overlap and forward (these are more typical for overlapping frames, and this workflow does not have that)
• Each group's output frames stored in set variables, and merged at end with image batch multi for the full video (instead of accumulating frames and forward to next groups). Might be a little less RAM heavy.
• ... plus all the small things from above (more accurate audio duration, better fps conditioning etc)<
• A bit simpler alternative than the "save each segment" wf.

(not really sure both workflows are needed, since the save intermediate files wf might be the better option for many/most. But since it used to be 2 variants.. They are quite similar though)

Plus made an excuse for a little rock video when testing ;-) 🤘🎸🎶

Hey RuneXX, I am in desperate need for your help, I need the the prompt for this video please if you still have it (the music style and the lyrics and the seed 😀 ).. I love this music and I want to create a song with same music. please if you still have it.

Hey RuneXX, I am in desperate need for your help, I need the the prompt for this video please if you still have it (the music style and the lyrics and the seed 😀 ).. I love this music and I want to create a song with same music. please if you still have it.

Will try search for the audio, ref images, and prompt i used ... and share if I can find it ;-)
Can be a nice starting point to try out the workflow i guess..

Hey RuneXX, I am in desperate need for your help, I need the the prompt for this video please if you still have it (the music style and the lyrics and the seed 😀 ).. I love this music and I want to create a song with same music. please if you still have it.

Will try search for the audio, ref images, and prompt i used ... and share if I can find it ;-)
Can be a nice starting point to try out the workflow i guess..

Thank you so much, the most important thing is the music style prompt please, I know it is too much to ask but I uploaded your song to Grok and asked to give me a prompt for the music and it failed :))