<div data-youtube-video><iframe width="640" height="480" allowfullscreen="true" autoplay="false" disablekbcontrols="false" enableiframeapi="false" endtime="0" ivloadpolicy="0" loop="false" modestbranding="false" origin playlist src="https://www.youtube.com/embed/ESU7isCNZww?rel=1" start="0"></iframe></div>This workflow is designed for LTX 2.3 single-person digital human voiceover generation, with IC Edit-style control and a strong focus on clean, subtitle-free output. Its main purpose is to take a reference portrait or character image, generate a stable talking-avatar video, and keep the final result clean for publishing without unwanted subtitles, captions, overlays, watermarks, or random text appearing on the screen.Compared with a normal image-to-video workflow, this setup is built specifically for single-character presentation. The goal is not to create chaotic movement or a dramatic action scene. The goal is controlled speaking performance: stable identity, stable camera framing, natural mouth motion, subtle facial expression, slight head movement, restrained body motion, and a clean final video that can be used as a digital host, AI presenter, product narrator, course explainer, or short-form talking avatar.The workflow uses LTX 2.3 as the core video generation route, with LTX video VAE, LTX audio-video latent logic, LTXVConditioning, LTXVPreprocess, image-to-video conditioning, motion-track-control IC LoRA, multiple SamplerCustomAdvanced stages, ManualSigmas, latent upscaling, tiled VAE decoding, and final video export. This makes the workflow more advanced than a basic one-pass digital human graph. It is designed to build the base motion first, then refine the latent result through additional passes so the final output looks more stable and usable.A key part of this workflow is the IC LoRA motion-track control route. For digital human generation, too much motion is often a problem. If the model freely changes the pose, face, background, clothing, or camera angle, the video becomes unusable as a talking-avatar clip. The IC LoRA control helps reduce random drift and keeps the character closer to the original visual identity while still allowing mouth movement, facial performance, and light natural motion.The workflow also uses a staged rendering structure. The first sampling stage establishes the base talking video from the input image and prompt conditions. Later stages continue from the generated latent result with lower sigma ranges, helping improve stability, detail, and texture without completely destroying the original portrait structure. This is especially important for single-person digital human videos because small face changes, broken hands, distorted mouth shapes, unstable eyes, or background flicker can quickly make the result feel fake.The “100% no subtitle” direction is also important. Many AI video outputs accidentally generate text, captions, watermarks, logo-like shapes, or subtitle bars, especially when the character is speaking. This workflow’s negative-control logic is designed to suppress subtitles, text on screen, overlays, watermarks, mismatched lip sync, robotic voice feeling, scene cuts, jump cuts, identity drift, extra people, deformed hands, extra fingers, and random style changes.This workflow is ideal for creators who want to build clean single-person digital human videos for YouTube, Bilibili, product introductions, AI course narration, social media explainers, virtual hosts, Civitai previews, and RunningHub demonstrations. If you want to see how LTX 2.3, IC Edit-style motion control, staged sampling, no-subtitle restrictions, latent refinement, and final digital human export work together, watch the full tutorial from the YouTube link above.⚙️ Try the Workflow Online👉 Workflow: <a target="_blank" rel="ugc" href="https://www.runninghub.ai/post/2054453027771297794?inviteCode=rh-v1111">https://www.runninghub.ai/post/2054453027771297794?inviteCode=rh-v1111</a>Open the link above to run the workflow directly online and view the generation results in real time.If the results meet your expectations, you can also deploy it locally for further customization.🎁 Fan Benefits: Register now to get 1000 points, plus 100 daily login points — enjoy 4090-level performance and 48 GB of powerful compute!📺 Bilibili Updates (Mainland China &amp; Asia-Pacific)If you are in Mainland China or the Asia-Pacific region, you can watch the video below for workflow demos and a detailed creative breakdown.📺 Bilibili Video: <a target="_blank" rel="ugc" href="https://www.bilibili.com/video/BV1za5y6FE7r/">https://www.bilibili.com/video/BV1za5y6FE7r/</a>I will continue updating model resources on Quark Drive:👉 <a target="_blank" rel="ugc" href="https://pan.quark.cn/s/20c6f6f8d87b">https://pan.quark.cn/s/20c6f6f8d87b</a>These resources are mainly prepared for local users, making creation and learning more convenient.⚙️ 在线体验工作流👉 工作流： <a target="_blank" rel="ugc" href="https://www.runninghub.ai/post/2054453027771297794?inviteCode=rh-v1111">https://www.runninghub.ai/post/2054453027771297794?inviteCode=rh-v1111</a>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： <a target="_blank" rel="ugc" href="https://www.bilibili.com/video/BV1za5y6FE7r/">https://www.bilibili.com/video/BV1za5y6FE7r/</a>我会在 夸克网盘 持续更新模型资源：👉 <a target="_blank" rel="ugc" href="https://pan.quark.cn/s/20c6f6f8d87b">https://pan.quark.cn/s/20c6f6f8d87b</a>这些资源主要面向本地用户，方便进行创作与学习。

Files with SHA256: ddd02e12dbffb6fa39164624d3aac8c3fa013cc99bb0f5795758f4ebe6d4fa1c