Kijai

/

LTX2.3_comfy

like 271

I2V & T2V Music Video Creator with Multi-Scenes and custom audio
On-the-fly music video creation with any input audio and images.
Creates a multi-scene video from your input images.

the video example is low resolution, so it can be even better if you run it at full HD res

Each scene can be timed correctly with a node for setting independent length for each scene.

For music creation: Ace Step XL (native workflow inside ComfyUI available). You can of course use any and all other music sources ;-)

For scene creation: Qwen Image Edit / FireRed Image Edit (native workflow inside ComfyUI available). With multi-scene lora, and multi-angles lora, creating the input images can be easy in ComfyUI.
Or you can use Nano Banana, Qwen Chat, ChatGPT etc. (For some music it might be more suitable with multiple singers, or illustration scenes etc. All up to you)

In the folder there is also two "helper" workflows. One for creating music with Ace Step XL, and one for creating new scenes from an input image with Qwen Image Edit. The Ace Step XL workflow has an LLM text generate optional part to help with describing music style/elements and lyrics creation using Kijai's great text generate node. Soon Gemma 4 will also be supported an it should work even better.

Inspired by @vrgamedevgirl84 and @MattHVisual that both have done some great music video related generations ;-)

Feel free to play around with it. Its a bit Work-in-Progress, so changes might come ;-)
(and will add a single-pass variant that can both generate new scene as well as extend a scene, and can use combo of both in same workflow)
https://huggingface.co/RuneXX/LTX-2.3-Workflows/tree/main/Music-Video-Creator

NB! This workflow might need a lot of ram, so if you are on a weaker computer, probably better to create part by part and edit it together in a video editor, instead of a workflow that does multiple videos in a row ;-)