Text-encoder weights (GGUF) for the LTX 2.5 inference path (engine25) of
Nz-Videomni (formerly Nz-LTX23), a plugin for the
video editor AviUtl2.
This file is a GGUF quantization of Lightricks' LTX-2.5-specific fine-tune of Gemma 4 (Google
DeepMind, Apache License 2.0), converted by Rootport using
Nz-GGUF-Converter-LTX23 (tool 1.2.0,
commit 905f348): 328 of 686 tensors are Q4_K, 2 are Q6_K, and the remaining 356 tensors (351 BF16
5 I8) are passed through unmodified.
Not compatible with general-purpose Gemma 4 — this checkpoint was fine-tuned exclusively to
serve as LTX 2.5's text encoder. Not compatible with the LTX 2.3 Gemma 3 text encoder
(Rootport/Nz-Gemma3-12B) either.
The video-generation weights (transformer, VAE, upscaler) are hosted separately at
Rootport/Nz-LTX25-weights.
License: subject to both the LTX-2.x Community License Agreement (LICENSE,
dated August 11, 2026, as the distribution license under which Lightricks released this file as
part of LTX-2.5) and the Apache License 2.0 (LICENSE-APACHE-2.0, as the
base model's license). See NOTICE.md for full attribution and obligations. Not an
official distribution of Google DeepMind or Lightricks Ltd., and not endorsed by either.