Add new model Scenema-Audio

[supermind-poster]

Could you add Scenema-Audio that was built on LTX 2.3 in Comfy? Seems it's very convinent speech generator that supports developed XML tags and works pretty well for generating speech and environment sounds. It could be usefull for LTX 2.3 generations to use output sounds as reference audio.
https://huggingface.co/ScenemaAI/scenema-audio

[farfromexact]

This probably needs a model integration or custom node rather than only adding the model name to ComfyUI core. The useful information for an implementation is the model architecture and weight format, tokenizer or text encoder, VAE or audio codec, sampler and scheduler assumptions, conditioning and XML tag handling, expected input and output types, VRAM and precision requirements, and the model license.

Because Scenema-Audio is built on LTX 2.3 and is intended to provide reference audio, the key compatibility question is whether it can consume the audio or latent types already used by the relevant LTX workflow, or whether an adapter is required. A small custom-node prototype with a reproducible workflow would be a good way to validate that before adding a maintained node to core.

The request would be stronger with the smallest working workflow, exact dependency versions, one input and output example, and a note about whether streaming, speech, and environment-sound generation are all required.

[supermind-poster]

They added custom node https://github.com/ScenemaAI/ComfyUI-ScenemaAudio by their own