← sources

epGen: open LoRA for LTX-2.5 that gives every sound in the video its own audio track (generation + separation, weights on HF)

source: reddit (reddit:StableDiffusion)published 2026-10-12 04:59ingested 2026-10-13 02:40

retained text

LTX-2.5 makes a video with one mixed soundtrack. Our LoRA also outputs one track per captioned sound, in the same run. Prompt "a jeep, its engine growling" and "a taxi, its horn honking" and you get the video, the full mix and each sound on its own track.

The same weights also separate the sounds of an existing video.

Walk into a scene with spatial audio (headphones):  
[https://sepgen.github.io](https://sepgen.github.io)

Two LoRAs for LTX-2.5 22B dev. Code and training configs included.

💻 [https://github.com/AviadDahan/SepGen](https://github.com/AviadDahan/SepGen)  
🤗 [https://huggingface.co/collections/AviadDahan/sepgen](https://huggingface.co/collections/AviadDahan/sepgen)

We're on today's Hugging Face Daily Papers. If you liked this work, an upvote there helps a lot!  [https://huggingface.co/papers/2610.11361](https://huggingface.co/papers/2610.11361)

Happy to answer questions about the setup or the training.

document fetched 2026-10-13 00:33 · canonical https://reddit.com/r/StableDiffusion/comments/1x3t05j/epgen_open_lora_for_ltx25_that_gives_every_sound