<h1 id="new-illuminate-ai-ltx-2.3-autonomous-vision-director-v1.2">NEW — ILLUMINATE AI LTX 2.3 AUTONOMOUS VISION DIRECTOR v1.2</h1><h2 id="complete-civitai-package-workflow-custom-nodes-and-installation">COMPLETE CIVITAI PACKAGE — WORKFLOW, CUSTOM NODES AND INSTALLATION</h2>ILLUMINATE AI v1.2 is a major quality and prompting update for the LTX 2.3 Autonomous Vision Director.The workflow analyzes the supplied reference image, understands the visible starting situation, creates a physically achievable chronological scene, writes the complete LTX 2.3 production prompt and validates it against the original image before video generation begins.The final validated prompt is displayed directly inside the workflow before it is sent to LTXDirector.<hr /><h2 id="new-in-v1.2">NEW IN v1.2</h2><h3 id="improved-image-quality">Improved Image Quality</h3>The complete render and refinement path has been reworked for noticeably cleaner output.Improvements include:<ul><li>cleaner hair structure during movement</li><li>finer skin, fabric and facial detail</li><li>reduced coarse and blocky motion textures</li><li>improved moving-edge quality</li><li>stronger Stage 2 detail reconstruction</li><li>native FP16 output without the previous downscale and bicubic enlargement chain</li></ul>Current quality configuration:<ul><li>Stage 1: 8 steps</li><li>Stage 1 CFG: 1.25</li><li>Spatial latent upscale: enabled</li><li>Stage 2: 4 steps</li><li>Stage 2 sigma endpoint: 0.40</li><li>Native FP16 output: 1.00</li><li>no third sampler required</li></ul><h3 id="more-accurate-autonomous-prompting">More Accurate Autonomous Prompting</h3>The autonomous Prompt Director has been expanded and refined.It now handles:<ul><li>duration-aware scene planning</li><li>more accurate participant and anatomy tracking</li><li>improved position and action recognition</li><li>better hand, limb and contact ownership</li><li>persistent anatomy consistency during occlusion</li><li>improved multi-participant continuity</li><li>primary and secondary action validation</li><li>more stable movement planning</li><li>stronger protection against morphing, anatomy loss and role changes</li></ul>The system remains image-guided and preserves the original scene whenever a complex transition cannot be completed safely.<hr /><h2 id="four-prompt-modes">FOUR PROMPT MODES</h2><h3 id="manual-prompt">Manual Prompt</h3>Uses your own finished prompt without an Ollama request.<h3 id="standard-vision">Standard Vision</h3>Qwen analyzes the reference image and creates a compatible cinematic or character-animation prompt.<h3 id="adult-assisted">Adult Assisted</h3>Uses the reference image together with a short written idea and expands it into a complete chronological scene.<h3 id="adult-full-auto">Adult Full Auto</h3>Creates the complete scene automatically from the reference image, selected duration and visible starting situation.Adult modes require confirmed adult and consenting participants.<hr /><h2 id="included-versions">INCLUDED VERSIONS</h2><h3 id="standard-recommended">STANDARD — RECOMMENDED</h3>The main release version.Use this version first for the best balance of:<ul><li>image quality</li><li>motion quality</li><li>anatomy consistency</li><li>render time</li><li>workflow stability</li></ul><h3 id="antighost-experimental">ANTIGHOST — EXPERIMENTAL</h3>An additional experimental version with optional temporal Anti-Ghost processing.This version is included for testing and comparison. It may improve ghosting or duplicated motion in some scenes, but it will not necessarily improve every video.The Anti-Ghost processing is disabled by default and can be activated for individual tests.<hr /><h1 id="important-new-custom-nodes-are-required">IMPORTANT — NEW CUSTOM NODES ARE REQUIRED</h1>The updated <code>ComfyUI-ILLUMINATE-AI-LTX</code> node pack included with v1.2 is mandatory.The newest workflow will not work correctly with an older ILLUMINATE node pack.The installation method has not changed.<h2 id="installation">Installation</h2><ol><li>Close ComfyUI completely.</li><li>Delete the existing folder:</li></ol><pre><code></code></pre><pre><code>ComfyUI\custom_nodes\ComfyUI-ILLUMINATE-AI-LTX</code></pre>Also remove the old standalone cleanup package if it is still installed:<pre><code></code></pre><pre><code>ComfyUI\custom_nodes\ComfyUI-Illuminate-Ollama-Cleanup</code></pre><ol><li> Copy the included folder: </li></ol><pre><code></code></pre><pre><code>02_CUSTOM_NODES\ComfyUI-ILLUMINATE-AI-LTX</code></pre>into:<pre><code></code></pre><pre><code>ComfyUI\custom_nodes\</code></pre>The final path must be:<pre><code></code></pre><pre><code>ComfyUI\custom_nodes\ComfyUI-ILLUMINATE-AI-LTX\__init__.py</code></pre>Do not create an additional nested folder.<ol><li> Load the included workflow and use: </li></ol><pre><code></code></pre><pre><code>ComfyUI Manager → Install Missing Custom Nodes</code></pre><ol><li> Restart ComfyUI completely. </li><li> Load the new v1.2 workflow JSON. </li></ol>A browser refresh with <code>Ctrl + F5</code> is recommended after updating the Custom Nodes.<hr /><h2 id="ollama-and-qwen">OLLAMA AND QWEN</h2>The autonomous Vision modes use:<pre><code></code></pre><pre><code>fredrezones55/Qwen3.5-Uncensored-HauhauCS-Aggressive:9b</code></pre>Install Ollama and run:<pre><code></code></pre><pre><code>ollama pull fredrezones55/Qwen3.5-Uncensored-HauhauCS-Aggressive:9b</code></pre>Ollama must remain running whenever Standard Vision, Adult Assisted or Adult Full Auto is used.Manual Prompt mode does not require Ollama.<hr /><h2 id="required-bat-flags">REQUIRED BAT FLAGS</h2>The included system check expects:<pre><code></code></pre><pre><code>--fp8_e4m3fn-text-enc --fast-disk</code></pre>Example:<pre><code></code></pre><pre><code>.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build --fp8_e4m3fn-text-enc --fast-disk</code></pre>A ready-made BAT example is included in the package.<hr /><h2 id="first-test">FIRST TEST</h2>For the first successful generation:<ul><li> load the same starting image into the LTXDirector and the Vision Reference Image node </li><li> use the STANDARD version first </li><li> leave optional LoRAs disabled </li><li> keep Anti-Ghost disabled </li><li> generate at 24 FPS </li><li> enable RIFE only after the normal 24 FPS output works </li><li> review the final generated prompt before LTX generation begins </li></ul><hr /><h2 id="package-content">PACKAGE CONTENT</h2>The download includes:<ul><li> v1.2 STANDARD workflow </li><li> v1.2 experimental ANTIGHOST workflow </li><li> updated mandatory ILLUMINATE Custom Nodes </li><li> Windows Portable BAT examples </li><li> Ollama/Qwen installation BAT </li><li> English installation guide </li><li> German quick-start guide </li><li> model and node checklist </li><li> troubleshooting information </li><li> release notes </li></ul>Models are not included in the ZIP.Please read <code>00_READ_ME_FIRST.txt</code> and the included installation guide before the first run.

ILLUMINATE AI – LTX 2.3 I2V AUTO-PROMPT Workflow

ILLUMINATE AI – LTX 2.3 Mainstream I2V WorkflowA carefully fine-tuned LTX 2.3 Image-to-Video workflow optimized for around 12 GB of VRAM.This version is focused on efficiency, stability, motion consistency, subject preservation, and reduced morphing while keeping generation fast and practical for mainstream GPUs.The workflow uses a two-pass setup to improve motion quality and maintain cleaner, more reliable results without excessive hardware demands. It is designed for one input image and one complete prompt to generate a continuous shot.Optimized for balanced 480p generation, lower memory usage, strong prompt understanding, and dependable image-to-video performance.Organized, ready to use, and built for users who want stable LTX 2.3 I2V results on lower-VRAM hardware.Custom Nodes NoticeThis workflow uses third-party custom nodes, including LTX-related nodes from Dasiwa and additional LTX workflow components such as LTX Director. Please make sure all required custom nodes are installed before running the workflow.Credits go to the original developers and node creators whose work made this workflow possible.Version Changes<ul><li>Added optional RIFE frame interpolation from 24 FPS to 48 FPS</li><li>Added a separate interpolated 48 FPS video output</li><li>Original 24 FPS output remains fully available</li><li>Audio stays synchronized when using the 48 FPS output</li><li>Interpolation is disabled by default</li><li>Added clear setup notes for correct 24 → 48 FPS usage</li><li>Added Low VRAM-friendly RIFE settings:<ul><li>Source FPS: 24</li><li>Target FPS: 48</li><li>Scale: 1.0</li><li>Batch Size: 12</li><li>FP16: Enabled</li></ul></li><li>Added separate filename prefixes for 24 FPS and 48 FPS outputs</li><li>Improved output organization and final workflow clarity</li></ul>Important:Enable both the RIFE interpolation node and the 48 FPS Video Combine node when using interpolation.The interpolated Video Combine must remain set to 48 FPS. Using 24 FPS with the interpolated frames will double the video duration and cause audio desynchronization.Interpolation improves motion smoothness but does not repair existing morphing, motion smearing or unstable anatomy.

Files with SHA256: f2f3c0b6c3f6dfd37ff0998c604b78909b873943ad3a440338cc0eb9ca26221a