# Inverting the Streaming-Diffusion Bottleneck: Video-Rate MLLM-Conditioned Edit Diffusion on a Consumer GPU

arXiv: https://arxiv.org/abs/2606.05981 · HTML: https://mlsanity.com/paper/2606.05981/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Multimodal distillation, one of 12 directions in Knowledge distillation.
- Builds on 48 of its 54 references that are in the map.
- First posted on arXiv 4 Jun 2026.

## Placed in

- Knowledge distillation / [Multimodal distillation](https://mlsanity.com/distill/multimodal-distillation/)

## Builds on (most cited first, 30 of 48)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [The Unreasonable Effectiveness of Deep Features as a Perceptual Metric](https://mlsanity.com/paper/1801.03924/) (2018)
- [Adding Conditional Control to Text-to-Image Diffusion Models](https://mlsanity.com/paper/2302.05543/) (2023)
- [Scaling Rectified Flow Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2403.03206/) (2024)
- [Qwen3-VL Technical Report](https://mlsanity.com/paper/2511.21631/) (2025)
- [Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets](https://mlsanity.com/paper/2311.15127/) (2023)
- [RAFT: Recurrent All-Pairs Field Transforms for Optical Flow](https://mlsanity.com/paper/2003.12039/) (2020)
- [Fast Inference from Transformers via Speculative Decoding](https://mlsanity.com/paper/2211.17192/) (2022)
- [AnimateDiff: Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuning](https://mlsanity.com/paper/2307.04725/) (2023)
- [The 2017 DAVIS Challenge on Video Object Segmentation](https://mlsanity.com/paper/1704.00675/) (2017)
- [Emerging Properties in Unified Multimodal Pretraining](https://mlsanity.com/paper/2505.14683/) (2025)
- [Adversarial Diffusion Distillation](https://mlsanity.com/paper/2311.17042/) (2023)
- [SDXL-Lightning: Progressive Adversarial Diffusion Distillation](https://mlsanity.com/paper/2402.13929/) (2024)
- [Learning Blind Video Temporal Consistency](https://mlsanity.com/paper/1808.00449/) (2018)
- [Rerender A Video: Zero-Shot Text-Guided Video-to-Video Translation](https://mlsanity.com/paper/2306.07954/) (2023)
- [LCM-LoRA: A Universal Stable-Diffusion Acceleration Module](https://mlsanity.com/paper/2311.05556/) (2023)
- [Parallel Sampling of Diffusion Models](https://mlsanity.com/paper/2305.16317/) (2023)
- [Adaptive Caching for Faster Video Generation with Diffusion Transformers](https://mlsanity.com/paper/2411.02397/) (2024)
- [DistriFusion: Distributed Parallel Inference for High-Resolution Diffusion Models](https://mlsanity.com/paper/2402.19481/) (2024)
- [FlowVid: Taming Imperfect Optical Flows for Consistent Video-to-Video Synthesis](https://mlsanity.com/paper/2312.17681/) (2023)
- [MotionStream: Real-Time Video Generation with Interactive Motion Controls](https://mlsanity.com/paper/2511.01266/) (2025)
- [SageAttention3: Microscaling FP4 Attention for Inference and An Exploration of 8-Bit Training](https://mlsanity.com/paper/2505.11594/) (2025)
- [StreamDiffusion: A Pipeline-level Solution for Real-time Interactive Generation](https://mlsanity.com/paper/2312.12491/) (2023)
- [InfiniPot-V: Memory-Constrained KV Cache Compression for Streaming Video Understanding](https://mlsanity.com/paper/2506.15745/) (2025)
- [OpenVE-3M: A Large-Scale High-Quality Dataset for Instruction-Guided Video Editing](https://mlsanity.com/paper/2512.07826/) (2025)
- [StreamDiT: Real-Time Streaming Text-to-Video Generation](https://mlsanity.com/paper/2507.03745/) (2025)
- [Looking Backward: Streaming Video-to-Video Translation with Feature Banks](https://mlsanity.com/paper/2405.15757/) (2024)
- [GPT4Video: A Unified Multimodal Large Language Model for lnstruction-Followed Understanding and Safety-Aware Generation](https://mlsanity.com/paper/2311.16511/) (2023)
- [Attention Is All You Need for KV Cache in Diffusion LLMs](https://mlsanity.com/paper/2510.14973/) (2025)
- [StreamDiffusionV2: A Streaming System for Dynamic and Interactive Video Generation](https://mlsanity.com/paper/2511.07399/) (2025)

## Abstract (arXiv)

Aggressive distillation of the diffusion U-Net inverts the per-frame bottleneck of real-time text-to-image pipelines: once the denoiser is a 4-step or 1-step distilled student, the text encoder becomes the critical path. This inversion is most acute in vision-aware edit diffusion, where the encoder is a multimodal large language model (MLLM). We study a 0.39B distilled edit U-Net paired with a 2.13B MLLM text encoder (Qwen3-VL) and present a streaming pipeline for this regime built on three mechanisms that keep the encoder off the denoiser's critical path rather than shrinking the encoder: asymmetric side-stream / main-stream CUDA pipelining with batched text-encoder amortisation, a compile-friendly ControlNet-LLLite reformulation that folds the whole U-Net + adapter stack into one fused graph, and a periodic conditioning-refresh schedule with a hook subset that amortises the per-frame conditioning cost. On a single consumer RTX 3090 Ti at 512x512 this sustains 27-30 fps over a 480-frame run; at the same operating point steady-state throughput scales to 55 fps on RTX 4090 and 74 fps on RTX 5090. This shows that once distillation is aggressive enough, further gains come from encoder-side systems work rather than further denoiser compression -- the opposite lever from the one the streaming-diffusion literature has optimised to date. We report video-rate streaming throughput, not interactive low latency, and locate our numbers against same-stack StreamDiffusion re-runs as systems context, not a superiority claim. The released oil-painting adapter generalises within in-clip noise to 19 unused DAVIS-2017 sequences and 15 non-DAVIS clips from seven sources; prompt-level generalisation to unseen styles is bounded and reported separately.