Search papers, labs, and topics across Lattice.
This paper introduces a series of KVAE tokenizers tailored for multimodal generative models, specifically KVAE-Audio, KVAE-3D, and KVAE-2D, which optimize the mapping of audio, video, and image inputs into compressed representations. By leveraging these tokenizers, the authors demonstrate significant improvements in both reconstruction and generation quality across various metrics, outperforming existing state-of-the-art open-source alternatives. The findings underscore the critical role of tokenization in enhancing the performance of latent diffusion models, particularly in text-conditioned generation tasks.
KVAE tokenizers achieve superior generative performance, setting a new benchmark for multimodal models in audio, image, and video synthesis.
Latent diffusion modeling (LDM), a prominent paradigm, utilizes tokenizers to map input signal to compressed representation. This dependency positions tokenizer as an integral part of generation process itself, since it affects learning speed, quality of synthesized samples and lay foundation for later applications. This report presents series of KVAE tokenizers for audio, image and video, all designed for subsequent text-conditioned generation: KVAE-Audio, a continuous full-band 48 kHz tokenizer with a 50 Hz latent of 64 channels; KVAE-3D -- two causal video tokenizers for 4x16x16 and 4x8x8 compression; KVAE-2D, an image model, compressing input by factor of 8 with 32 channels. We demonstrate that reconstruction (PSNR, LPIPS, PESQ, etc.) and generation results on objective (Frechet Distance, CLIP score, CLAP score, etc.) and subjective (side-by-side evaluation) metrics matches or surpasses frontier opensource tokenizers, such as VAEs from Wan-2.2, HunyuanVideo-1.5, FLUX.2, MovieGen, StableAudio and MMAudio. Considering difficulty of development, we share with community training details, model selection method and ablation on design choices. The code is publicly available at https://github.com/kandinskylab/kvae and https://github.com/kandinskylab/kvae-audio.
