LTX-2.3 Foley LoRA: Realistic Video-to-Audio Sound Effects for Lightricks' Video Model

news

FuzzPuppy releases LTX-2.3 Foley LoRA, a video-to-audio LoRA that adds realistic, visually synchronized sound effects to LTX-2.3 generated videos without background music overlay.

FuzzPuppy has released **LTX-2.3 Foley LoRA** (HuggingFace), a LoRA adapter for Lightricks' LTX-2.3 that generates realistic, visually synchronized sound effects for generated video content — without adding background music or score.

Overview

LTX-2.3 Foley is a video-to-audio LoRA for Lightricks' LTX-2.3 video generation model. While LTX-2.3 natively generates video with accompanying audio, it often defaults to background music or rhythmic soundtrack material. The Foley LoRA redirects the model toward producing real-world sound effects that match the visual content — footsteps, doors closing, kitchen sounds, sports impacts, and more.

The name "Foley" comes from the filmmaking technique of recording custom sound effects in sync with on-screen action.

How It Works

The LoRA is applied on top of LTX-2.3's video-to-audio capabilities. Users control the output through:

  • LoRA multiplier: Recommended range 1–3 (higher values suppress music more aggressively)
  • Descriptive prompts: Describe the visible action, followed by "No speech is present. No music is present."
  • Negative prompting: Exclude music, melody, singing, and score elements

Example prompt:

A barista uses an espresso machine to steam milk. No speech is present. No music is present.

Compatibility

The Foley LoRA is designed for LTX-2.3 base or distilled video-to-audio models. It is available as a single .safetensors file (ltx-2.3-foley-400-steps.safetensors, 400 training steps) and can be loaded via standard LoRA loading workflows in ComfyUI or diffusers.

Availability

The LoRA weights are released under the LTX Community License Agreement on HuggingFace:

LTX-2.3 Foley LoRA: Realistic Video-to-Audio Sound Effects for Lightricks' Video Model | ComfyUI Wiki