Watch the full video first if you want to understand how this LTX 2.3 storyboard director workflow works in practice. The video shows how character, action, camera movement, and scene control can be organized into one directed video pipeline, instead of relying on a single loose text prompt.

This ComfyUI workflow is designed for LTX 2.3 director-style video generation, focusing on integrated control of characters, actions, scenes, and cinematic shot structure. Its main purpose is to make LTX video generation behave more like a storyboard-driven production workflow. Instead of writing one long prompt and hoping the model understands the full scene, this workflow uses LTXDirectorGuide and PromptRelay-style control to divide the video into a more structured director pipeline.

The workflow is built around ltx-2.3-22b-dev-dare-ties-distilled-1.1.safetensors as the main video checkpoint. It also uses a Gemma3 fp8 text encoder, LTX Audio VAE, LTXVConditioning, LTXDirectorGuide, LTX2_NAG, OmniNFT LoRA, latent upscaling, AV latent concatenation and separation, tiled VAE decoding, and final video output. The graph also includes model-chain logic for stability, quality control, optional LoRA equipment, and a three-stage rendering route.

The core value of this workflow is director control. LTXDirectorGuide allows the video generation process to receive more structured guide data, so the model can follow planned shot logic more clearly. This is useful when the video needs controlled character identity, specific body action, camera movement, scene relationship, lighting direction, and staged visual rhythm. In normal text-to-video workflows, these elements are often mixed together inside one prompt. In this workflow, they are separated into a cleaner director-style control system.

The generation process is divided into three major stages. The first stage performs the initial director sampling. It establishes the base motion, camera structure, character placement, and scene direction. The second stage performs latent-space upscaling and refinement with the LTX 2.3 x2 spatial upscaler. This stage improves structure and prepares the video for higher-quality output. The third stage applies final light sampling and HD polish, using another independent Director Guide path and final sampler route.

The workflow also uses LTX2_NAG and a universal negative prompt system to suppress common video problems such as flicker, frame jitter, identity drift, warped anatomy, subtitles, UI overlays, watermarks, random text, distorted audio, and unstable motion. OmniNFT is also connected as part of the director model chain, helping the workflow maintain stronger identity, visual style, and reward-aligned cinematic behavior.

Compared with ordinary LTX 2.3 text-to-video workflows, this director workflow is more suitable for planned shots. It is not only about generating a beautiful clip; it is about controlling who appears, what they do, where they are, how the camera moves, and how the shot progresses. This makes it useful for AI short films, character story clips, cinematic previews, action staging, MV fragments, product-style narrative shots, Bilibili demonstrations, YouTube showcases, RunningHub releases, and Civitai workflow publishing.

Main features:

• LTX 2.3 director-style video workflow

• Character, action, and scene integrated control

• LTXDirectorGuide / PromptRelay control structure

• LTX 2.3 distilled 1.1 checkpoint route

• Gemma3 fp8 text encoder

• LTX Audio VAE support

• OmniNFT LoRA identity and style guidance

• LTX2_NAG quality stabilization

• Universal negative prompt system

• Three-stage director rendering pipeline

• Stage 1 initial director sampling

• Stage 2 latent upscaling and refinement

• Stage 3 final HD polish

• LTX 2.3 x2 latent upscaler support

• VAEDecodeTiled and final video output

Suggested workflow:

Start by writing the video like a storyboard. Define the main character, action, environment, camera movement, lighting, and scene rhythm before rendering. Use the director prompt structure to make the character behavior and camera direction clear. Run the first stage first and check whether the scene blocking, action, and camera logic are correct. If the first stage is wrong, fix the director prompt before entering refinement. After the base shot works, continue into the second-stage latent upscale and refinement route. Use the third stage only after the motion and composition are already stable. If the video drifts, simplify the prompt, reduce conflicting actions, and keep the negative prompt active.

⚙️ RunningHub Workflow

Try the workflow online right now — no installation required.
👉 Workflow: https://www.runninghub.ai/post/2061524156398788609?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/BV1nVVr6QEd8/

☕ 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/2061524156398788609?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/BV1nVVr6QEd8/

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。