Joy-LTX 2.5 — One Take or a Seamless Multishot, Picture and Sound in One Pass (ComfyUI)

v1.1.0 - adds the PLUS canvases ( JoyLTX25_Take_PLUS , JoyLTX25_Multishot_PLUS ): a LoRA stack (four slots on the DiT; on a DEV merge the official distilled LoRA at 0.4-0.6), Refs by Name (cast characters from input/joyecho_refs/ / photo folders - one photo per character, and from a character's second shot on their own rendered face and voice carry into every later shot), Prompts from File (your own shot list, single file or a folder walked by an incrementing episode index - queue once per file), and audio export (flac). The loader is now JoyLTX Load Model (takes .safetensors or .gguf, resolves a moved file by name). The plain Take / Multishot canvases are unchanged apart from the loader and the v2 model names. File: Joy-LTX_2.5_v1.1.0.zip . Models on Hugging Face: GGUF · comfy-native · DEV merges . Default DiT in the canvases: LTX25dist-echoVid-070T30-v2-DiT-comfy-int8.safetensors - pick your own file in the loader if your card is smaller.