Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2601.03233 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914236802793472 |
|---|---|
| author | HaCohen, Yoav Brazowski, Benny Chiprut, Nisan Bitterman, Yaki Kvochko, Andrew Berkowitz, Avishai Shalem, Daniel Lifschitz, Daphna Moshe, Dudu Porat, Eitan Richardson, Eitan Shiran, Guy Chachy, Itay Chetboun, Jonathan Finkelson, Michael Kupchick, Michael Zabari, Nir Guetta, Nitzan Kotler, Noa Bibi, Ofir Gordon, Ori Panet, Poriya Benita, Roi Armon, Shahar Kulikov, Victor Inger, Yaron Shiftan, Yonatan Melumian, Zeev Farbman, Zeev |
| author_facet | HaCohen, Yoav Brazowski, Benny Chiprut, Nisan Bitterman, Yaki Kvochko, Andrew Berkowitz, Avishai Shalem, Daniel Lifschitz, Daphna Moshe, Dudu Porat, Eitan Richardson, Eitan Shiran, Guy Chachy, Itay Chetboun, Jonathan Finkelson, Michael Kupchick, Michael Zabari, Nir Guetta, Nitzan Kotler, Noa Bibi, Ofir Gordon, Ori Panet, Poriya Benita, Roi Armon, Shahar Kulikov, Victor Inger, Yaron Shiftan, Yonatan Melumian, Zeev Farbman, Zeev |
| contents | Recent text-to-video diffusion models can generate compelling video sequences, yet they remain silent -- missing the semantic, emotional, and atmospheric cues that audio provides. We introduce LTX-2, an open-source foundational model capable of generating high-quality, temporally synchronized audiovisual content in a unified manner. LTX-2 consists of an asymmetric dual-stream transformer with a 14B-parameter video stream and a 5B-parameter audio stream, coupled through bidirectional audio-video cross-attention layers with temporal positional embeddings and cross-modality AdaLN for shared timestep conditioning. This architecture enables efficient training and inference of a unified audiovisual model while allocating more capacity for video generation than audio generation. We employ a multilingual text encoder for broader prompt understanding and introduce a modality-aware classifier-free guidance (modality-CFG) mechanism for improved audiovisual alignment and controllability. Beyond generating speech, LTX-2 produces rich, coherent audio tracks that follow the characters, environment, style, and emotion of each scene -- complete with natural background and foley elements. In our evaluations, the model achieves state-of-the-art audiovisual quality and prompt adherence among open-source systems, while delivering results comparable to proprietary models at a fraction of their computational cost and inference time. All model weights and code are publicly released. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_03233 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | LTX-2: Efficient Joint Audio-Visual Foundation Model HaCohen, Yoav Brazowski, Benny Chiprut, Nisan Bitterman, Yaki Kvochko, Andrew Berkowitz, Avishai Shalem, Daniel Lifschitz, Daphna Moshe, Dudu Porat, Eitan Richardson, Eitan Shiran, Guy Chachy, Itay Chetboun, Jonathan Finkelson, Michael Kupchick, Michael Zabari, Nir Guetta, Nitzan Kotler, Noa Bibi, Ofir Gordon, Ori Panet, Poriya Benita, Roi Armon, Shahar Kulikov, Victor Inger, Yaron Shiftan, Yonatan Melumian, Zeev Farbman, Zeev Computer Vision and Pattern Recognition Recent text-to-video diffusion models can generate compelling video sequences, yet they remain silent -- missing the semantic, emotional, and atmospheric cues that audio provides. We introduce LTX-2, an open-source foundational model capable of generating high-quality, temporally synchronized audiovisual content in a unified manner. LTX-2 consists of an asymmetric dual-stream transformer with a 14B-parameter video stream and a 5B-parameter audio stream, coupled through bidirectional audio-video cross-attention layers with temporal positional embeddings and cross-modality AdaLN for shared timestep conditioning. This architecture enables efficient training and inference of a unified audiovisual model while allocating more capacity for video generation than audio generation. We employ a multilingual text encoder for broader prompt understanding and introduce a modality-aware classifier-free guidance (modality-CFG) mechanism for improved audiovisual alignment and controllability. Beyond generating speech, LTX-2 produces rich, coherent audio tracks that follow the characters, environment, style, and emotion of each scene -- complete with natural background and foley elements. In our evaluations, the model achieves state-of-the-art audiovisual quality and prompt adherence among open-source systems, while delivering results comparable to proprietary models at a fraction of their computational cost and inference time. All model weights and code are publicly released. |
| title | LTX-2: Efficient Joint Audio-Visual Foundation Model |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2601.03233 |