Updated: May 9, 2026
character
This ComfyUI workflow is designed for FlashVSR v1.1 video upscaling, video restoration, frame enhancement, and high-resolution video export. The main purpose of this workflow is to take an input video, extract its frames, process those frames through FlashVSR v1.1, and then recombine the enhanced frames back into a finished video while keeping the original audio track.
Unlike an image-only upscaling workflow, this graph is built for video. It handles video frame loading, frame preprocessing, AI-based video super-resolution, audio passthrough, and final MP4 export inside one clean ComfyUI pipeline. This makes it suitable for creators who want to improve video clarity, upscale AI-generated clips, enhance low-resolution footage, clean soft animation output, or prepare sharper videos for YouTube, Bilibili, RunningHub, Civitai, and social media publishing.
The workflow begins with VHS_LoadVideo. This node loads the source video into ComfyUI and outputs the video frames, frame count, audio track, and video information. In the included setup, the workflow keeps the original video timing behavior, loads every frame, does not skip the first frames, and does not force a new input frame rate or forced input size. This keeps the source video structure simple and avoids unnecessary preprocessing changes before the upscale stage.
One important advantage of this workflow is that the audio is preserved. The audio output from VHS_LoadVideo is passed into VHS_VideoCombine at the final stage. This means the enhanced video can keep the original sound, music, narration, voice, or effects track. For AI video creators, this is important because many clips already include voiceover, lip-sync audio, sound design, background music, or edited narration. The workflow improves the visual frames without forcing the user to reattach audio manually in another program.
After video loading, the workflow uses LayerUtility: ImageScaleByAspectRatio V2 to prepare the frames before FlashVSR processing. In the included setup, the aspect ratio is set to original, the fit mode is fill, the scaling method is Lanczos, the output is rounded to a multiple of 8, and the longest side is controlled around 960 pixels before the FlashVSR stage. This preprocessing step helps control the working resolution and makes the input more suitable for stable video super-resolution.
This frame preparation stage is useful because video upscaling can quickly become heavy. If the input frames are too large, memory usage and processing time may increase significantly. If the frames are too small or poorly scaled, the final output may not have enough detail. The aspect-ratio scaling stage provides a practical balance: keep the original layout, use a clean scaler, and prepare the frames in a controlled size before AI enhancement.
The core enhancement module is FlashVSRNode. In this workflow, the model is set to FlashVSR-v1.1, the mode is set to full, and the scale value is set to 3. This means the workflow is configured for a strong video upscaling pass. The FlashVSR node receives the prepared video frames and outputs enhanced frames for final recombination.
FlashVSR is used here as a video super-resolution tool rather than a simple still-image upscaler. Video upscaling needs to consider frame-to-frame consistency. A normal image upscaler can improve single frames, but it may create inconsistent details from frame to frame, which can cause flickering in video. A video super-resolution workflow is more suitable for improving moving footage because it is designed around frame sequences.
The included FlashVSR settings enable tiled_vae and tiled_dit. These tiled settings are important for practical video processing because video frames can consume a lot of VRAM, especially when scaling up. Tiling helps manage memory usage and makes the workflow more suitable for larger frames or longer clips. The workflow also includes seed control with randomization, allowing different restoration behaviors when testing multiple outputs.
The final stage uses VHS_VideoCombine to rebuild the enhanced frames into a video file. In the included setup, the output format is video/h264-mp4, the pixel format is yuv420p, CRF is set to 19, metadata saving is enabled, save_output is enabled, pingpong is disabled, and the output frame rate is set around 25 fps. This makes the workflow suitable for direct video export in a widely compatible MP4 format.
The use of H.264 MP4 output is practical for publishing. It can be uploaded to YouTube, Bilibili, social media platforms, RunningHub previews, and general web players. The CRF value gives a reasonable balance between quality and file size. For higher-quality exports, users can lower the CRF value. For smaller files, they can increase it slightly.
This workflow is especially useful for AI video creators. Many AI video generation models output clips that look good in motion but are still limited in resolution, sharpness, or local detail. FlashVSR v1.1 can be used as a final enhancement stage after image-to-video, text-to-video, lip-sync video, digital human video, anime video, product video, or cinematic short clip generation. It helps make the final video cleaner and more suitable for publishing.
It is also useful for restoring older or compressed footage. If a video has soft edges, low resolution, weak texture, or compression blur, this workflow can provide a stronger high-resolution output. The result still depends on the source quality, but the workflow is a practical first step for video enhancement inside ComfyUI.
Main features:
- FlashVSR v1.1 video upscaling workflow
- Video input through VHS_LoadVideo
- Frame extraction and audio passthrough
- Aspect-ratio-preserving frame preparation
- Lanczos preprocessing
- Longest-side control before AI upscaling
- FlashVSRNode enhancement using FlashVSR-v1.1
- Full mode processing
- 3x video upscale configuration
- Tiled VAE support
- Tiled DiT support
- H.264 MP4 export through VHS_VideoCombine
- Original audio preserved in final output
- Suitable for AI video, animation, digital human clips, and general footage enhancement
Recommended use cases:
AI video upscaling, text-to-video enhancement, image-to-video post-processing, digital human video sharpening, lip-sync video cleanup, anime video upscale, product video enhancement, social media video preparation, YouTube upload preparation, Bilibili video publishing, RunningHub workflow showcase, Civitai demo video enhancement, low-resolution footage restoration, compressed video cleanup, and final AIGC video polishing.
Suggested workflow:
Start by loading your source video into VHS_LoadVideo. Use a video with stable motion and acceptable source quality. FlashVSR can improve clarity and resolution, but if the source is extremely damaged, heavily compressed, or full of motion artifacts, the result may still need additional cleanup.
Check the video loading settings before running the workflow. If you want to process the whole video, keep frame_load_cap at 0. If you only want a fast test, set a small frame limit first. This is useful because video upscaling can be much heavier than image upscaling.
Use select_every_nth carefully. The included setup processes every frame, which is best for normal video quality. If you skip frames, the output may become less smooth or require additional frame interpolation later. For final output, processing every frame is usually recommended.
Use the aspect-ratio scaling node to control working size. The included setup keeps the original aspect ratio and scales the longest side to around 960 before FlashVSR processing. This is a practical starting point for balancing quality and speed. If your source is very small, this gives FlashVSR a cleaner working input. If your source is already large, you may need to reduce the preprocessing size to avoid excessive VRAM use.
Run FlashVSR v1.1 in full mode for stronger enhancement. The included scale value is 3, which creates a significant resolution increase. This is useful for turning softer video into a sharper final version. If your hardware struggles with memory, keep tiled_vae and tiled_dit enabled.
After processing, VHS_VideoCombine recombines the enhanced frames into an MP4 video. The original audio is connected into the combine node, so the final output keeps the sound from the source video. This is useful for narration clips, music videos, lip-sync videos, and any video where the audio timing must remain intact.
Use CRF control based on your publishing target. The included CRF value around 19 is a good general-purpose quality setting. Lower CRF values create larger files with higher quality. Higher CRF values create smaller files but may introduce more compression artifacts.
For short clips, test at full settings. For long videos, test a short segment first. Video upscaling can take time, so it is better to confirm the visual style, sharpness, and output format before processing a long video. Once the short test looks correct, run the full clip.
When evaluating the result, look for more than sharpness. Check frame consistency, motion stability, edge quality, face stability, texture quality, and whether the enhancement introduces flicker or over-sharpening. Good video upscaling should improve clarity without making motion unstable.
For AI-generated videos, this workflow is best used as a final enhancement stage. Generate your video first, then use FlashVSR v1.1 to upscale and polish it. For digital human or lip-sync videos, make sure the mouth and face do not become overly sharpened. For anime clips, check line stability and edge cleanliness. For product videos, inspect logos, text, surface reflections, and fine material details.
This workflow is designed as a simple but practical FlashVSR v1.1 video upscale pipeline for ComfyUI users. It combines video loading, frame preprocessing, FlashVSR enhancement, audio preservation, and MP4 export into one graph. It is especially useful for creators who want to improve AI-generated videos before publishing them to YouTube, Bilibili, RunningHub, Civitai, or social media.
🎥 YouTube Video Tutorial
Want to know what this workflow actually does and how to start fast?
This video explains what the tool is, how to launch the workflow instantly, and shares my core design logic — no local setup, no complicated environment.
Everything starts directly on RunningHub, so you can experience it in action first.
👉 YouTube Tutorial: https://youtu.be/OjsHOyPtF0s
Before you begin, I recommend watching the video thoroughly — getting the full context helps you understand the tool faster and avoid common detours.
⚙️ RunningHub Workflow
Try the workflow online right now — no installation required.
👉 Workflow: https://www.runninghub.ai/post/2018310765513547778/?inviteCode=rh-v1111
If the results meet your expectations, you can later deploy it locally for customization.
🎁 Fan Benefits: Register to get 1000 points + daily login 100 points — enjoy 4090 performance and 48 GB super power!
📺 Bilibili Updates (Mainland China & Asia-Pacific)
If you’re in the Asia-Pacific region, you can watch the video below to see the workflow demonstration and creative breakdown.
📺 Bilibili Video: https://www.bilibili.com/video/BV1mwFLzTELL/
☕ Support Me on Ko-fi
If you find my content helpful and want to support future creations, you can buy me a coffee ☕.
Every bit of support helps me keep creating — just like a spark that can ignite a blazing flame.
👉 Ko-fi: https://ko-fi.com/aiksk
💼 Business Contact
For collaboration or inquiries, please contact aiksk95 on WeChat.
🎥 YouTube 视频教程
想了解这个工作流到底是怎样的工具，以及如何快速启动？
视频主要介绍 工具定位、快速启动方法 和 我的构筑思路。
我们会直接在 RunningHub 上进行演示，让你第一时间看到实际效果。
👉 YouTube 教程： https://youtu.be/OjsHOyPtF0s
开始前建议尽量完整地观看视频 —— 把握整体思路会更快上手，也能少走常见弯路。
⚙️ 在线体验工作流
现在就可以在线体验，无需安装。
👉 工作流： https://www.runninghub.ai/post/2018310765513547778/?inviteCode=rh-v1111
打开上方链接即可直接运行该工作流，实时查看生成效果。
如果觉得效果理想，你也可以在本地进行自定义部署。
🎁 粉丝福利： 注册即送 1000 积分，每日登录 100 积分，畅玩 4090 体验 48 G 超级性能！
📺 Bilibili 更新（中国大陆及南亚太地区）
如果你在中国大陆或南亚太地区，可以通过下方视频查看该工作流的实测效果与构思讲解。
📺 B站视频： https://www.bilibili.com/video/BV1mwFLzTELL/
我会在 夸克网盘 持续更新模型资源：
👉 https://pan.quark.cn/s/20c6f6f8d87b
这些资源主要面向本地用户，方便进行创作与学习。