RuneXX

/

LTX-2.3-Workflows

like 757

Wav used is from this video

https://packaged-media.redd.it/5d433ftqndrg1/pb/m2-res_720p.mp4?m=DASHPlaylist.mpd&c=wh_ben_en&var=sgpssan&v=1&e=1774738800&s=6af54c37564ea683f31aa7af514956d982c67339

Running Windows. I think it has something to do with that. Anyone running windows and has this working?

I'm on windows too ;-) will check if anything has changed or something broke

Are you running portable or desktop?

I got it working. I had to replace all the files. Not just the nodes_lt.py

https://github.com/Comfy-Org/ComfyUI/issues/13194

ah. You are on the desktop version of ComfyUI I guess then.
The desktop version the updates rolls out a bit slower, but the ID-Lora update will come eventually ;-)

Hey RuneXX, do you know, is this ID-LoRA for audio only or is it both audio and visual ID?

Great workflows RuneXX, thanks a lot. Which of your workflows do you actually find better for this purpose? LTX-2.3_-I2V_T2V_Basic_ID-Lora_reference_audio.json, LTX-2.3-I2V_T2V_Talking_Avatar(voice_clone_with_Fish-Audio-Pro).json or LTX-2.3_-I2V_T2V_Talking_Avatar(voice_clone_with_Qwen-TTS).json?

I have tried the ID-Lora and Fish Audio. I find that ID-Lora creates better cohesion with better facial motor skills. Fish Audio has slightly better voice quality and is less robotic, but the motor skills are significantly worse.

By the way, do you know a way to improve mouth and tooth quality?

Hey RuneXX, do you know, is this ID-LoRA for audio only or is it both audio and visual ID?

Its an audio only lora, to make consistent voice across multiple videos

By the way, do you know a way to improve mouth and tooth quality?

Could try more steps, or higher resolution.
And yes the ID lora is trained on expressions (face and voice), so it might be more expressive .. true
But you could make up for that with Fish Audio and the prompt.. probably. With prompts like "she talks with an expressive face and gesticulate as she talks with an emotional tone..." or something else that suits your input ;-) LTX loves a good prompt, for how to act .. and sequence of acting.

Hi @RuneXX

I'm having a problem with LTX-2.3_-_I2V_T2V_Basic_ID-Lora_reference_audio.json

I'm using some custom music / speech via the load audio node but it doesn't seem to be getting picked up properly in the final video

I'm basically playing a song with some lyrics, i want the person to dance to the song and sing the vocals, is it possible to do it using this workflow please?

Currently i get an audio output unrelated to the song track so i'm not sure if i'm doing something wrong or prompting it wrong?

Also if i try to run in single pass mode (disabling the two options) i then get an error in the LTXVSeperateAVLatent node - ComfyUI\execution.py", line 534, in execute
output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)

not sure if its related to my issues?

Please can you help? 😊

basically playing a song with some lyrics, i want the person to dance to the song and sing the vocals, is it possible to do it using this workflow please?

For that you dont want to use the ID Lora workflow. The ID lora workflow is more a "voice clone" of sorts. Where it takes the voice in the reference audio, as input for a new dialog via prompting

When you want to use the audio input as is (speak, music, etc etc), you want to use a regular custom audio workflow (without ID-Lora).
You can use the "Basic Custom Audio" workflow here: https://huggingface.co/RuneXX/LTX-2.3-Workflows/tree/main/Custom-Audio
That should do the job ;-)

--

Try the basic custom audio above first. And if you want to get more "advanced" you can try the Prompt-Relay workflow with custom audio.
What it does is basically let you prompt different segments of your video, making it easier to control the "story" of the video output... Think "first she does this, then she does this, and finally this"
https://huggingface.co/RuneXX/LTX-2.3-Workflows/tree/main/Movie-Maker/Prompt-Relay-Custom-Audio
(there is also a multi-image version there, that takes multiple images to make a longer "story", but maybe start with the non multi-image first.. . they are easier)

You can of course get by without Prompt-Relay, and get much the same result, its just a bit easier, and stronger influence of your segmented prompt when using Prompt Relay

The "Basic Custom Audio" one works great for this thank you