🚀 MrXin LTX 2.3 I2V EROS V3

This is the improved version of the LTX 2.3 Eros workflow. The main goal of V3 was to solve the biggest complaint from V2: almost no in motion between the First Pass and the Final Pass.

V3 Major Upgrades:

• Significantly more motion preserved in the Final Pass (much closer to the lively First Pass)

• New Model-based Upscaler added (nmkdSiaxCX) – works on cloud servers and local machines

• Option to completely disable the Final Pass and only save the First Pass (faster generation)

• Support for alternative base model (distilled transformer) – you can now easily switch between the original Eros checkpoint and the distilled model

• Cleaner workflow with better switches and memory cleanup nodes

• Custom Height and Width selection.

✅ What You Get

• Strong, natural movement in both First and Final Pass

• High-quality 24 FPS final video with baked audio

• Optional Model Upscaler (great for cloud users)

• Easy toggle to skip Final Pass entirely

• Two live previews + clean output folders

📥 Required Models + Direct Download Links

• Checkpoint → ltx2310eros_beta.safetensors

• Distilled Model → ltx-2.3-22b-distilled_transformer_only_fp8_input_scaled_v3.safetensors

• Text Encoder → gemma_3_12B_it_fp8_e4m3fn.safetensors

• Text Projection → ltx-2.3_text_projection_bf16.safetensors

• Video VAE → LTX23_video_vae_bf16.safetensors

• Audio VAE → LTX23_audio_vae_bf16.safetensors

• Preview VAE → taeltx2_3.safetensors

• Spatial Upscaler → ltx-2.3-spatial-upscaler-x2-1.1.safetensors

• Model Upscaler → nmkdSiaxCX_200k.safetensors
  

📁 Folder Structure

ComfyUI/
├── models/
│   ├── checkpoints/
│   │   └── ltx2310eros_beta.safetensors
│   │
│   ├── diffusion_models/
│   │   └── ltx-2.3-22b-distilled_transformer_only_fp8_input_scaled_v3.safetensors
│   │
│   ├── text_encoders/
│   │   └── gemma_3_12B_it_fp8_e4m3fn.safetensors
│   │
│   ├── clip/
│   │   └── ltx-2.3_text_projection_bf16.safetensors
│   │
│   ├── VAE/
│   │   ├── LTX23_video_vae_bf16.safetensors
│   │   ├── LTX23_audio_vae_bf16.safetensors
│   │   └── taeltx2_3.safetensors
│   │
│   ├── latent_upscale_models/
│   │   └── ltx-2.3-spatial-upscaler-x2-1.1.safetensors
│   │
│   ├── upscale_models/
│   │   └── nmkdSiaxCX_200k.safetensors
│   │
│   └── Lora/
│       └── ltx-2.3-22b-distilled-lora-dynamic_fro09_avg_rank_105_bf16.safetensors

Pro Tip: You can now choose between the original Eros checkpoint or the distilled model using the switches.

🛠️ Custom nodes and How to Use remain the same as V2 (just load the new V3 JSON).

💡 Pro Tips (V3)

• Want maximum motion? Lower the strength in LTXVImgToVideoInplace (Final Pass) to 0.6 or 0.55

• Want faster renders? Disable the Final Pass group and only save the First Pass

• Cloud users: Use the new Model Upscaler instead of RTX Super Resolution

This version finally delivers what most people wanted: strong movement that actually survives the final pass.

Enjoy V3! 🔥

— MrXin (March 2026)


------------------------------------------------------------------------------------------------------

🚀 Image to Video V2 – What’s New & Improved (March 2026)

• Built-in VIDEO EDITOR (bypassable) → fully automatic 48 FPS final video

  • RTX Video Super Resolution (1.5x + ULTRA quality)

  • RIFE Interpolation (2x FPS)

  • Two-stage workflow (generate base → then upscale) to keep VRAM usage low

• Newer & better models

  • Spatial Upscaler upgraded to v1.1 (ltx-2.3-spatial-upscaler-x2-1.1)

  • Lighter & faster text encoder (gemma_3_12B_it_fp8_e4m3fn)

• Expanded LoRA suite (more explicit power)

  • LTX2.3 10Eros

  • DR34ML4Y

  • LTX 2/2.3 [I2V] NSFW (+furry)

  • Cum Shot

  • LTX 2.3 - POV Paizuri - Titfuck I2V

  • Orgasm

  • LTX-2.3 Deepthroat

• Improved final decode with LTXV Spatio-Temporal Tiled VAE Decode (sharper details, fewer artifacts)

• New trigger-word notes added directly in the workflow (Deepthroat, Creampie animation, Stomach bulge, Wet Dr1pp, etc.)

• Cleaner output folders (/I2V/ for base + /Editor/ for 48 FPS final)

• Extra stability & VRAM optimizations (VRAM_Debug, Fast Groups Bypasser, better batch management)

Result: Same rock-solid 12GB VRAM performance, but noticeably smoother motion, much better explicit details, higher final FPS (48) and overall higher visual quality.

V1 → V2 upgrade is 100% worth it if you want that extra polish and modern 48 FPS look without extra hassle.

📥 Required Models + Direct Download Links Main files (click to download directly):

• Checkpoint → ltx2310eros_beta.safetensors

• Text Encoder (fp8 – super light) → gemma_3_12B_it_fp8_e4m3fn.safetensors

• Spatial Upscaler → ltx-2.3-spatial-upscaler-x2-1.1.safetensors

• Preview VAE → taeltx2_3.safetensors

• Distilled LoRA → ltx-2.3-22b-distilled-lora-dynamic_fro09_avg_rank_105_bf16.safetensors
  
  Folder structure

ComfyUI/
├── models/
│   ├── checkpoints/            ← ltx2310eros_beta.safetensors
│   ├── clip/                   ← gemma_3_12B_it_fp8_e4m3fn.safetensors
│   ├── VAE/                    ← taeltx2_3.safetensors
│   ├── latent_upscale_models/  ← ltx-2.3-spatial-upscaler-x2-1.1.safetensors
│   └── Lora/                   ← ltx-2.3-22b-distilled-lora-dynamic_fro09_avg_rank_105_bf16.safetensors

🛠️ Custom Nodes Required

Install via ComfyUI Manager (same as V1 + a couple extras for the editor):

• ComfyUI-LTXVideo + KJNodes (latest)

• rgthree-comfy

• ComfyUI-easy-use

• ComfyUI-VideoHelperSuite (VHS)

• ComfyUI-mxToolkit

• GACLove/ComfyUI-VFI (for RIFE)

• comfyui_nvidia_rtx_nodes (for RTX Super Resolution)

------------------------------------------------------------------------------------------------------

🚀 MrXin LTX 2.3 I2V EROS V1

This is a fully-featured, production-ready ComfyUI workflow for LTX 2.3 (Eros Beta) that turns a single image + text prompt into a 20-second cinematic video with synchronized audio, realistic motion, and optional explicit animation.

Specially optimized for low VRAM users (tested and runs smoothly on RTX 5080 12GB + 32GB RAM thanks to:

• LTXVChunkFeedForward (4 chunks)

• Distilled LoRA (0.6)

• fp8 text encoder

• 3-stage progressive sampling (First → Second → Final Pass)

• Empty audio latent + normalized audio output

✅ What You Get

• Smooth 24 FPS MP4 with baked audio (moans, groans, impacts)

• Three live preview windows: First Pass • Second Pass • Final Video

• Automatic image resize (default longer side 1280)

• Final video saved with audio

📥 Required Models + Direct Download Links

Main files:

• Checkpoint → ltx2310eros_beta.safetensors

• Text Encoder (fp8 – only ~11 GB) → gemma_3_12B_it_heretic_fp8_e4m3fn.safetensors

• Spatial Upscaler → ltx-2.3-spatial-upscaler-x2-1.0.safetensors

• Preview VAE → taeltx2_3.safetensors

• Distilled LoRA → ltx-2.3-22b-distilled-lora-dynamic_fro09_avg_rank_105_bf16.safetensors

Folder structure

ComfyUI/
├── models/
│   ├── checkpoints/          ← ltx2310eros_beta.safetensors
│   ├── clip/                 ← gemma_3_12B_it_heretic_fp8_e4m3fn.safetensors
│   ├── VAE/                  ← taeltx2_3.safetensors
│   ├── latent_upscale_models/← ltx-2.3-spatial-upscaler-x2-1.0.safetensors
│   └── Lora/                 ← ltx-2.3-22b-distilled-lora-dynamic_fro09_avg_rank_105_bf16.safetensors

🛠️ Custom Nodes Required

Install via ComfyUI Manager:

• ComfyUI-LTXVideo + KJNodes updates

• rgthree-comfy

• ComfyUI-easy-use

• ComfyUI-VideoHelperSuite (VHS)

• ComfyUI-mxToolkit

Restart after install.

🎮 How to Use

1. Load the JSON workflow

2. Drag your starting image into the Load Image node

3. Edit the Positive Prompt

4. Adjust sliders:

   • Video Length (seconds): 20 (default)

   • Longer Side: 1280 (lower it if you OOM)

5. Hit Queue Prompt

6. Watch all three previews update live

7. Final video with audio is automatically saved to output/MrXin/LTX2.3/

🔍 Workflow Structure

• Load Models

• Load LoRAs + Distilled LoRA

• Load Image + Settings

• Prompts

• Processing Preview (the 3-pass engine)

• First / Second / Final Pass Previews + Output

💡 Pro Tips

• Works great on 12GB VRAM at 1280 longer side + 25 seconds

• For lower VRAM: set Longer Side to 1024 and Length to 10

• The Chunk FeedForward node is key for stability – leave at 4 chunks
  

✅ Quick OOM Fix Guide

If you’re getting Out of Memory errors or weird faces in the final pass, follow these steps in order:

1. Update or Fresh Install ComfyUI + Install Missing Custom Nodes

• Download and run the Easy Installer: 👉 https://github.com/Tavris1/ComfyUI-Easy-Install

• Or watch this short video: 👉 https://www.youtube.com/watch?v=CgLL5aoEX-s&t=1041s

After updating/installing:

• Open ComfyUI → go to Manager → click Install Missing Custom Nodes

• Restart ComfyUI completely.

2. Add the Low VRAM Flags

1. Go to your ComfyUI folder.

2. Find the file ComfyUI.bat.

3. Right-click → Edit (open with Notepad).

4. Find the line that has ComfyUI\main.py .

5. Add the flags at the end so it looks like this:

bat

 ComfyUI\main.py --lowvram --reserve-vram 6 --preview-method none --disable-xformers --disable-smart-memory

1. Save and launch ComfyUI using this .bat file.

3. Increase Windows Swap File (Virtual Memory) This helps when VRAM still spikes:

1. Right-click This PC → Properties → Advanced system settings.

2. Click Settings under Performance → Advanced tab → Change (under Virtual memory).

3. Uncheck “Automatically manage paging file size for all drives”.

4. Select your main drive (usually C:) → choose Custom size.

5. Set:

   • Initial size (MB): 49152

   • Maximum size (MB): 98304

6. Click Set → OK → restart your PC.

After doing all three steps, reload the workflow and test with Longer Side 1024 and Video Length 20 seconds first. This combination fixes OOM for almost everyone on 12GB cards.

This workflow delivers excellent motion, audio sync, and timing accuracy while staying lightweight. Just load once, drop your image, tweak prompt & sliders, and generate.

Ready to post on CivitAI! 🔥

— MrXin (March 2026)

Disclaimer:

This workflow is provided for entertainment, artistic, and creative purposes only.

It may not be used for any illegal, harmful, non-consensual, or malicious activities.

Please use it responsibly and respect all applicable laws and ethical guidelines.