Watch the full video first if you want to understand how this LTX2.3 3.5 intelligent video aspect expansion workflow works in practice. The video shows how an existing video can be expanded into a wider or different frame format while preserving the original motion, subject, timing, and visual continuity as much as possible.
This ComfyUI workflow is designed for LTX2.3 3.5 video aspect-ratio expansion and video outpainting. Its main purpose is to take an existing video and intelligently extend the canvas beyond the original frame. This is useful when creators need to convert footage into a different publishing format, such as expanding a vertical video into a horizontal layout, extending a cropped shot, creating more space around the subject, or preparing a video for platform-specific aspect ratios.
The workflow is built around ltx-2.3-22b-dev-dare-merged-distilled-1.1.safetensors as the main LTX2.3 video model. The text encoding route uses gemma_3_12B_it_fp8_e4m3fn.safetensors through the LTX AV text encoder loader. The workflow also loads the LTX audio VAE from the same checkpoint, allowing the graph to preserve the audio-video latent structure during the expansion process.
The key module is the LTX2.3 outpaint LoRA route. The workflow uses ltx-2.3-22b-ic-lora-outpaint.safetensors through LTXICLoRALoader, then injects the video guide through LTXAddVideoICLoRAGuide. This gives the model a stronger understanding of the original video content while allowing it to generate new visual areas outside the source frame. The goal is not to redesign the whole video, but to extend the missing frame area in a way that matches the original scene.
The input video is preprocessed through LTXVPreprocess, then resized and aligned through image-resize and size-reading nodes. The workflow uses GetImageSize and EmptyLTXVLatentVideo to build the target latent video size. In the provided setup, the latent video is configured around a wider 960×544 frame with 121 frames, making it suitable for horizontal video expansion tests.
The conditioning route uses LTXVConditioning at 24fps. The positive and negative prompt inputs define what the expanded area should look like and what should be avoided. The negative prompt suppresses unwanted game-like visuals, ugly results, static output, or other style conflicts. This helps the expanded region blend better with the original video rather than looking like a separate generated image pasted onto the side.
After the first latent construction, LTXVSeparateAVLatent separates video and audio latent components. LTXVCropGuides helps manage guide alignment for the expanded canvas. LTXVConcatAVLatent then recombines the video and audio latent structure before sampling. The main sampling pass uses RandomNoise, CFGGuider, ManualSigmas, KSamplerSelect, and SamplerCustomAdvanced to generate the outpainted result.
The decoding and final assembly section uses VAEDecodeTiled for memory-friendly frame decoding. The workflow also includes color correction and switch branches, including an optional route for darker scenes where normal outpainting may not work well. ImageConcanate is used to combine visual regions, and VHS_VideoCombine exports the final video in a RunningHub-safe H264 MP4 format.
Main features:
LTX2.3 3.5 video aspect expansion workflow
Intelligent video outpainting
Expands video canvas beyond the source frame
Preserves original subject and motion continuity
ltx-2.3-22b-dev-dare-merged-distilled-1.1.safetensors support
Gemma3 FP8 text encoder support
LTX Audio VAE support
LTX2.3 outpaint IC LoRA support
LTXAddVideoICLoRAGuide guide injection
LTXVPreprocess video preparation
ImageResizeKJv2 and GetImageSize sizing route
EmptyLTXVLatentVideo target canvas creation
LTXVCropGuides alignment control
Audio-video latent separation and recombination
ManualSigmas sampling schedule
SamplerCustomAdvanced generation
VAEDecodeTiled memory-friendly decoding
Color correction branch
Dark-scene compatibility switch
ImageConcanate frame combination
VHS_VideoCombine H264 MP4 final export
Suggested workflow:
Upload a video with a clear subject and stable framing first. Decide whether you want to expand the video horizontally, vertically, or into a new platform format. Keep the prompt focused on describing the missing environment around the original frame rather than rewriting the entire video. If the expanded sides look inconsistent, simplify the prompt and describe the surrounding scene more clearly. If the original subject changes too much, strengthen the preservation wording and reduce aggressive style terms. For dark scenes, use the dark-scene branch if the normal outpainting route does not blend cleanly. Start with short tests before processing longer clips.
⚙️ RunningHub Workflow
Try the workflow online right now — no installation required.
👉 Workflow: https://www.runninghub.ai/post/2067822318637895681?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/BV1xw7F6XE7K/
☕ 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.
⚙️打开下方链接即可在线体验，无需安装。
👉 工作流： https://www.runninghub.ai/post/2067822318637895681?inviteCode=rh-v1111
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/BV1xw7F6XE7K/
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。