We are excited to announce the release of LongCat-Video-Avatar 1.5, an upgraded open-source framework that prioritizes extreme empirical optimization and production-readiness for audio-driven human video generation. Built upon the LongCat-Video foundation model, v1.5 delivers highly stable, commercial-grade avatar video synthesis supporting native tasks including Audio-Text-to-Video (AT2V), Audio-Text-Image-to-Video (ATI2V), and Video Continuation, with seamless compatibility for both single-stream and multi-stream audio inputs.
Key Features
🌟 Upgraded Audio Encoder (Whisper-Large):: Replaces Wav2Vec2 with Whisper-Large, yielding significantly smoother and more natural lip dynamics.
🌟 Production-Ready Stability: Achieves accurate lip-synchronization, full-body temporal stability, and robust long-video generation with strict identity consistency.
🌟 Stylized Domain Generalization: Robustly generalizes to anime, animals, and complex real-world conditions such as multi-person interactions and object handling.
🌟 Efficient 8-Step Inference: Advanced DMD2-based step distillation accelerates inference to 8 NFE, balancing cost-effective serving with exceptional visual fidelity.
LongCat-Video-Avatar 1.5 supports diverse application scenarios including broadcasting, acting, singing, e-commerce marketing, multi-person conversation, animation, and animal characters.
📊 Human Evaluation
We introduce a comprehensive human evaluation benchmark specifically tailored for audio-driven digital human generation. The benchmark encompasses 6 application scenarios (News Broadcasting, Knowledge Education, Daily Life, Entertainment, Singing, Commercial Promotion), 2 languages (Chinese/English), and 2 visual styles (Realistic/Animated), yielding a total of 508 image-audio source pairs. Evaluation Methodology:(1)Subjective Track: 770 crowdsourced evaluators rated each generated video on a 1–5 human-likeness scale, yielding 13,240 judgments. (2) Objective Track: 10 domain experts conducted structured quality analysis across four dimensions: Physical Rationality, Harmony (Audio-Visual Coordination), Temporal Stability, and Identity Consistency.
The results are in the following figure: (a) Expert-level objective quality evaluation across four dimensions (b)Subjective human-likeness comparison with leading commercial models.
LongCat-Video-Avatar
💡 Quick Start
Clone the repo
shell
1git clone --single-branch --branch main https://github.com/meituan-longcat/LongCat-Video
2cd LongCat-Video
Lip synchronization accuracy: Audio CFG works optimally between 3–5. Increase the audio CFG value for better synchronization.
Prompt Enhancement: Longer, more descriptive prompts yield better consistency and naturalness than short ones. We recommend including rich details such as character appearance, actions, and scene context (e.g., "A young woman with long black hair is speaking and smiling, wearing a white blouse, sitting in a bright café") for best results.
Mitigate repeated actions: Setting the reference image index（--ref_img_index, default to 10） between 0 and 24 ensures better consistency; setting it to 30 helps reduce repeated actions. Additionally, increasing the mask frame range (--mask_frame_range, default to 3) can further help mitigate repeated actions, but excessively large values may introduce artifacts.
Super resolution: Our model is compatible with both 480P and 720P, which can be controlled via --resolution.
Dual-Audio Modes: Merge mode (set audio_type to para) requires two audio clips of equal length, and the resulting audio is obtained by summing the two clips; Concatenation mode (set audio_type to add) does not require equal-length inputs, and the resulting audio is formed by sequentially concatenating the two clips with silence padding for any gaps, where by default person1 speaks first and person2 speaks afterward.
Model versions:--model_type avatar-v1.0 uses wav2vec2 audio encoder (default); --model_type avatar-v1.5 uses Whisper-large-v3 audio encoder for better lip sync quality.
Distillation mode: Add --use_distill to enable distillation sampling (fewer steps, faster inference). This is required when using --model_type avatar-v1.5.
INT8 quantization: Add --use_int8 to load the INT8 quantized DiT model for reduced VRAM usage. Only supported with --model_type avatar-v1.5.
The model weights are released under the MIT License.
Any contributions to this repository are licensed under the MIT License, unless otherwise stated. This license does not grant any rights to use Meituan trademarks or patents.
This model has not been specifically designed or comprehensively evaluated for every possible downstream application.
Developers should take into account the known limitations of large language models, including performance variations across different languages, and carefully assess accuracy, safety, and fairness before deploying the model in sensitive or high-risk scenarios.
It is the responsibility of developers and downstream users to understand and comply with all applicable laws and regulations relevant to their use case, including but not limited to data protection, privacy, and content safety requirements.
Nothing in this Model Card should be interpreted as altering or restricting the terms of the MIT License under which the model is released.
📖 Citation
We kindly encourage citation of our work if you find it useful.
