# Aero Realtime: Fully Aligned Input-Output Streams for Low-Latency Streaming Multimodal Generation

arXiv: https://arxiv.org/abs/2608.08469 · HTML: https://mlsanity.com/paper/2608.08469/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Audio and omni-modal token compression, one of 12 directions in Efficient LLM inference and serving.
- Builds on 19 of its 28 references that are in the map.
- First posted on arXiv 9 Aug 2026.

## Placed in

- Efficient LLM inference and serving / [Audio and omni-modal token compression](https://mlsanity.com/inference/audio-and-omni-modal-token-compression/)

## Builds on (most cited first, 19 of 19)

- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [LLaVA-OneVision: Easy Visual Task Transfer](https://mlsanity.com/paper/2408.03326/) (2024)
- [How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suites](https://mlsanity.com/paper/2404.16821/) (2024)
- [LongVU: Spatiotemporal Adaptive Compression for Long Video-Language Understanding](https://mlsanity.com/paper/2410.17434/) (2024)
- [DeepSpeed Ulysses: System Optimizations for Enabling Training of Extreme Long Sequence Transformer Models](https://mlsanity.com/paper/2309.14509/) (2023)
- [VideoLLM-online: Online Video Large Language Model for Streaming Video](https://mlsanity.com/paper/2406.11816/) (2024)
- [Liger Kernel: Efficient Triton Kernels for LLM Training](https://mlsanity.com/paper/2410.10989/) (2024)
- [Dispider: Enabling Video LLMs with Active Real-Time Interaction via Disentangled Perception, Decision, and Reaction](https://mlsanity.com/paper/2501.03218/) (2025)
- [TimeChat-Online: 80% Visual Tokens are Naturally Redundant in Streaming Videos](https://mlsanity.com/paper/2504.17343/) (2025)
- [StreamingVLM: Real-Time Understanding for Infinite Video Streams](https://mlsanity.com/paper/2510.09608/) (2025)
- [Flash-VStream: Efficient Real-Time Understanding for Long Video Streams](https://mlsanity.com/paper/2506.23825/) (2025)
- [StreamForest: Efficient Online Video Understanding with Persistent Event Memory](https://mlsanity.com/paper/2509.24871/) (2025)
- [StreamBridge: Turning Your Offline Video Large Language Model into a Proactive Streaming Assistant](https://mlsanity.com/paper/2505.05467/) (2025)
- [HERMES: KV Cache as Hierarchical Memory for Efficient Streaming Video Understanding](https://mlsanity.com/paper/2601.14724/) (2026)
- [AURA: Always-On Understanding and Real-Time Assistance via Video Streams](https://mlsanity.com/paper/2604.04184/) (2026)
- [VideoChat3: Fully Open Video MLLM for Efficient and Generalist Video Understanding](https://mlsanity.com/paper/2607.14935/) (2026)
- [Mage-VL: An Efficient Codec-Native Streaming Multimodal Foundation Model](https://mlsanity.com/paper/2607.24904/) (2026)
- [Vision-Speech Models: Teaching Speech Models to Converse about Images](https://mlsanity.com/paper/2503.15633/) (2025)
- [MOSS-Video-Preview: Toward Real-Time Video Understanding via Cross-Attention](https://mlsanity.com/paper/2606.07639/) (2026)

## Built on by (most cited first, 1 of 1)

- [OmniCapBench: A Deep-Structured Evaluation Framework for Fine-Grained Audio-Visual Captioning](https://mlsanity.com/paper/2610.12458/) (2026)

## Abstract (arXiv)

Existing streaming multimodal models process observations incrementally but still follow a turn-based prefill-then-decode pattern, making them non-duplex: new observations cannot naturally enter an active generation stream. Proactive alternatives use micro-turn polling or external response gates, which fragment continuous interaction, decouple response timing from language generation, and complicate KV-cache-friendly serving. We introduce Aero Realtime, a 4B streaming multimodal model with a duplex architecture for realtime generation. Aero Realtime aligns video, audio, and textual output on a shared temporal grid, where each approximately 80-ms audio slot predicts either a lexical token or a silence token. This allows input and output to advance together, enabling one autoregressive objective to learn both when to respond and what to generate. During inference, Aero Realtime appends only the newest multimodal slot, carries forward the previous output state, and reuses the KV cache for efficient incremental execution. We further provide a complete training and serving recipe, including realtime QA construction, slot-aligned supervision, hardware-aware distributed training, and resumable inference. On four NVIDIA A6000 workstation GPUs, Aero Realtime maintains 84-ms median and 173-ms P95 processing lag over 20 minutes of a continuously streamed video, remaining within 200~ms of the source timeline. These results demonstrate the feasibility of fully aligned input-output modeling for duplex, proactive, and hardware-aligned multimodal interaction.