# 🎬 H3 ErosLite: Minimalist 2-Pass Video Workflow &amp; Technical Guide&gt; **Author:** 4lth3rego &gt; **Base Model:** MiniMax H3 (ErosLite NVFP4 Turbo) &gt; **Target Framework:** ComfyUI &gt; **Pipeline:** 2-Pass Hybrid Generation (Base Turbo + 3D Latent Upscale + Cinema Sharpen)---## 🌟 1. Overview &amp; Architectural PhilosophyGenerating high-definition AI video on consumer hardware traditionally faces two severe bottlenecks:1. **Extreme VRAM Consumption:** High-resolution video diffusion natively requires massive memory buffers, often causing Out-Of-Memory (OOM) errors on 16 GB GPUs.2. **Facial Distortion in Upscaling:** Standard spatial or latent upscalers frequently hallucinate micro-textures, altering facial identity and introducing temporal flickering between frames.The **H3 ErosLite 2-Pass Workflow** solves both challenges through an engineered, modular pipeline:* **Stage 1 (Base Motion Generation):** Generates fluid, temporally coherent video at **0.4 MP (~864×480)** using **8 Turbo steps** at lightning speed.* **Stage 2 (3D Latent Upscaling):** Uses <code>MMH3UltimateUpscale</code> with a 3D latent diffusion model and a 1-step solver at a strictly calibrated denoise of **0.18**, upscaling to **0.9 – 1.0 MP (~1344×768)** with zero facial morphing.* **Stage 3 (Cinema Sharpening):** Applies a sub-second edge-refinement pass that brings out cinematic film clarity without artificial grain.* **Canvas Modularization:** The entire pipeline is organized into **5 clean Subgraphs**, reducing 60+ cluttered nodes into an intuitive, elegant workspace.---## 🧩 2. Detailed Breakdown of Workflow Subgraphs### 📦 Subgraph 1: H3 ErosLite Model Loader* **Primary Role:** Manages model weight ingestion and VRAM offloading.* **Checkpoint:** Loads <code>H3_ErosLite.safetensors</code> (14.25 GiB). Edge layers (blocks 0, 1, 47, 48, 49) are preserved in lossless **BF16**, while intermediate linear layers are quantized to **NVFP4 (FP4-E2M1 with FP8 block scales)**.* **Text &amp; Conditioning:** Loads <code>qwen3vl_32b_heretic_minimax_h3_nvfp4.safetensors</code> via <code>CLIPLoader</code>.* **VAEs:** Separately loads the Video VAE <code>minimax_h3_video_vae_fp16.safetensors</code>) and Audio VAE <code>minimax_h3_audio_vae_fp32.safetensors</code>).* **VRAM Footprint:** Fits comfortably on 16 GB GPUs (RTX 4080, RTX 5060 Ti, RTX 4070 Ti, RTX 3090).---### ⏱️ Subgraph 2: Duration to Frame Length* **Primary Role:** Enforces mathematical compatibility with MiniMax H3's temporal compression.* **The <code>17n + 5</code> Rule:** MiniMax H3's 3D VAE requires video frames to strictly conform to the formula: $$\text{Frames} = \max(5, \text{round}(t \times 24)) + \left(5 - (\max(5, \text{round}(t \times 24)) \pmod{17})\right) \pmod{17}$$* **Usage:** Enter duration in seconds (e.g. <code>5.0s</code>). The subgraph automatically computes **124 frames** (at 24 fps), preventing VAE slicing errors.---### 📐 Subgraph 3: Resolution Standardizer* **Primary Role:** Standardizes input geometry across different aspect ratios and generation passes.* **Core Functions:** * **Aspect Ratio Selection:** Select standard presets <code>16:9 widescreen 1344x768</code>, <code>9:16 vertical</code>, <code>1:1 square</code>, <code>5:8 portrait</code>, etc.) or pass original input dimensions. * **Pass 1 Sizing (0.4 MP):** Scales input to ~400,000 pixels (e.g., 864×480) strictly rounded to multiples of 32 for the base motion pass. * **Pass 2 Sizing (0.9 – 1.0 MP):** Computes the proportional upscale target (e.g., 1344×768) and directly drives the 3D upscaler parameters. * **Mode Switch:** Toggle <code>1</code> for Original Image Aspect Ratio, or <code>2</code> for Aspect Ratio Override / Resize.---### ✍️ Subgraph 4: H3 Promptor (Multimodal Engine)* **Primary Role:** Translates and structures creative concepts into MiniMax H3's native multimodal prompt syntax.* **Internal Stages:** 1. **🌐 Translation:** <code>GoogleTranslateTextNode</code> translates non-English inputs to clean English. 2. **👁️ Vision Feature Extraction:** <code>H3_Vision_Analyzer</code> analyzes the initial frame image <code>image_0</code>) to extract character identity, clothing, lighting, and palette. 3. **🧹 VRAM Purge:** Triggers <code>LayerUtility: PurgeVRAM V2</code> to free memory before diffusion starts. 4. **✍️ Multimodal Structuring:** <code>H3_Promptor</code> synthesizes the vision context and creative description into official MiniMax tags: * <code>&lt;d&gt;[Language] ... &lt;/d&gt;</code> * <code>integrated_multimodal_description:</code> * <code>overall_soundscape:</code> * <code>non_diegetic_music:</code> 5. **👁️ Live Preview:** The generated prompt is routed to a <code>ShowText</code> node for immediate inspection.---### 🔍 Subgraph 5: MMH3 2-Pass Latent 3D Upscaler* **Primary Role:** High-resolution spatial and micro-texture refinement.* **Engine:** <code>MMH3UltimateUpscale</code> paired with <code>minimax_h3_latent_upscaler_3d_bf16.safetensors</code>.* **Identity Preservation Setting:** Calibrated to **1 step** with <code>beta57</code> scheduler at **denoise = 0.18**. This guarantees that skin pores, hair strands, and lighting reflections are crisply added without altering facial anatomy or character identity.* **Cinema Sharpening:** Output frames pass through <code>ImageSharpen</code> (Radius: 1, Sigma: 0.8, Strength: 0.35) to remove digital softness and deliver filmic texture.---## 🚀 3. Step-by-Step User Guide### Mode A: Image-to-Video (I2V) — Recommended1. Drag and drop your source image into *<code>1. Start Frame</code>**.2. Set your desired duration in *<code>⏱️ Duration</code>** (default: <code>5.0s</code>).3. In *<code>✍️ H3 Promptor</code>**, type a description of the desired action (e.g. "A woman smiling and turning her head towards the camera on a windy beach").4. Leave *<code>2. End Frame</code>** empty.5. Click **Queue Prompt**.---### Mode B: First &amp; Last Frame (FL2V)1. Load your beginning image into *<code>1. Start Frame</code>**.2. Load your target finishing image into *<code>2. End Frame</code>**.3. In *<code>✍️ H3 Promptor</code>**, set <code>Task Type</code> to <code>FL2VA</code> or describe the transitional trajectory.4. Click **Queue Prompt**. The model will interpolate fluid motion connecting both keyframes.---### Mode C: Reference-to-Video (Ref2V)1. Load your reference character/style image into *<code>3. Reference Image</code>**.2. In **Group 4**, disconnect the latent from <code>MiniMaxH3ImageToVideo</code> and connect the latent output from <code>MiniMaxH3ReferenceToVideo</code> (Node 10) to the Sampler.3. Click **Queue Prompt**.---## ⚙️ 4. Recommended Parameter Reference Table| Parameter | Pass 1 (Base Generation) | Pass 2 (Latent 3D Upscale) | Notes || :--- | :--- | :--- | :--- || **Resolution** | ~864×480 (0.4 MP) | ~1344×768 (0.9 – 1.0 MP) | Driven automatically by Subgraph 3 || **Sampling Steps** | 8 Turbo Steps | 1 Step | Fast 8-step base + 1-step upscale || **Sampler** | <code>euler</code> | <code>er_sde</code> or <code>sa_solver</code> | Ultra-stable temporal coherence || **Scheduler** | <code>beta57</code> | <code>beta57</code> | Calibrated MiniMax curve || **Denoise** | <code>1.00</code> | *<code>0.18 – 0.20</code>** | **Do not exceed 0.25** to avoid face morphing || **Frame Rate** | 24 fps | 24 fps | Standard cinema frame rate || **Sharpening** | None | Radius: 1, Sigma: 0.8, Strength: 0.35 | Sub-second filmic enhancement |---## 💡 5. Best Practices &amp; Pro Tips* **Preserving Facial Likeness:** Keep Pass 2 denoise strictly between <code>0.16</code> and <code>0.20</code>. A denoise of <code>0.50</code> will rewrite facial features and destroy identity.* **Audio Synchronization:** The Audio VAE synthesizes a synchronized soundscape directly from the multimodal prompt. Use headphones to preview ambient audio and spatial sound effects.* **Prompting Secret:** Let <code>H3 Promptor</code> do the heavy lifting. Avoid pasting giant walls of raw tags; instead, describe natural camera movement (*"slow tracking shot"*, "medium close-up") and physical dynamics (*"hair blowing gently in the wind"*).---## 📥 6. Installation &amp; Files1. Place the model checkpoint in: <code>ComfyUI/models/diffusion_models/H3_ErosLite.safetensors</code>2. Place the CLIP model in: <code>ComfyUI/models/clip/qwen3vl_32b_heretic_minimax_h3_nvfp4.safetensors</code>3. Place the VAEs in: <code>ComfyUI/models/vae/minimax_h3_video_vae_fp16.safetensors</code> <code>ComfyUI/models/vae/minimax_h3_audio_vae_fp32.safetensors</code>4. Drag and drop the workflow file <code>H3_ErosLite_2Pass_Clean_Workflow.json</code> into ComfyUI.

H3 ErosLite - 2-Pass Minimalist Video Workflow

# 🎬 H3 ErosLite: Minimalist 2-Pass Video Workflow &amp; Technical Guide&gt; **Author:** 4lth3rego &gt; **Base Model:** MiniMax H3 (ErosLite NVFP4 Turbo) &gt; **Target Framework:** ComfyUI &gt; **Pipeline:** 2-Pass Hybrid Generation (Base Turbo + 3D Latent Upscale + Cinema Sharpen)---## 🌟 1. Overview &amp; Architectural PhilosophyGenerating high-definition AI video on consumer hardware traditionally faces two severe bottlenecks:1. **Extreme VRAM Consumption:** High-resolution video diffusion natively requires massive memory buffers, often causing Out-Of-Memory (OOM) errors on 16 GB GPUs.2. **Facial Distortion in Upscaling:** Standard spatial or latent upscalers frequently hallucinate micro-textures, altering facial identity and introducing temporal flickering between frames.The **H3 ErosLite 2-Pass Workflow** solves both challenges through an engineered, modular pipeline:* **Stage 1 (Base Motion Generation):** Generates fluid, temporally coherent video at **0.4 MP (~864×480)** using **8 Turbo steps** at lightning speed.* **Stage 2 (3D Latent Upscaling):** Uses <code>MMH3UltimateUpscale</code> with a 3D latent diffusion model and a 1-step solver at a strictly calibrated denoise of **0.18**, upscaling to **0.9 – 1.0 MP (~1344×768)** with zero facial morphing.* **Stage 3 (Cinema Sharpening):** Applies a sub-second edge-refinement pass that brings out cinematic film clarity without artificial grain.* **Canvas Modularization:** The entire pipeline is organized into **5 clean Subgraphs**, reducing 60+ cluttered nodes into an intuitive, elegant workspace.---## 🧩 2. Detailed Breakdown of Workflow Subgraphs### 📦 Subgraph 1: H3 ErosLite Model Loader* **Primary Role:** Manages model weight ingestion and VRAM offloading.* **Checkpoint:** Loads <code>H3_ErosLite.safetensors</code> (14.25 GiB). Edge layers (blocks 0, 1, 47, 48, 49) are preserved in lossless **BF16**, while intermediate linear layers are quantized to **NVFP4 (FP4-E2M1 with FP8 block scales)**.* **Text &amp; Conditioning:** Loads <code>qwen3vl_32b_heretic_minimax_h3_nvfp4.safetensors</code> via <code>CLIPLoader</code>.* **VAEs:** Separately loads the Video VAE <code>minimax_h3_video_vae_fp16.safetensors</code>) and Audio VAE <code>minimax_h3_audio_vae_fp32.safetensors</code>).* **VRAM Footprint:** Fits comfortably on 16 GB GPUs (RTX 4080, RTX 5060 Ti, RTX 4070 Ti, RTX 3090).---### ⏱️ Subgraph 2: Duration to Frame Length* **Primary Role:** Enforces mathematical compatibility with MiniMax H3's temporal compression.* **The <code>17n + 5</code> Rule:** MiniMax H3's 3D VAE requires video frames to strictly conform to the formula: $$\text{Frames} = \max(5, \text{round}(t \times 24)) + \left(5 - (\max(5, \text{round}(t \times 24)) \pmod{17})\right) \pmod{17}$$* **Usage:** Enter duration in seconds (e.g. <code>5.0s</code>). The subgraph automatically computes **124 frames** (at 24 fps), preventing VAE slicing errors.---### 📐 Subgraph 3: Resolution Standardizer* **Primary Role:** Standardizes input geometry across different aspect ratios and generation passes.* **Core Functions:** * **Aspect Ratio Selection:** Select standard presets <code>16:9 widescreen 1344x768</code>, <code>9:16 vertical</code>, <code>1:1 square</code>, <code>5:8 portrait</code>, etc.) or pass original input dimensions. * **Pass 1 Sizing (0.4 MP):** Scales input to ~400,000 pixels (e.g., 864×480) strictly rounded to multiples of 32 for the base motion pass. * **Pass 2 Sizing (0.9 – 1.0 MP):** Computes the proportional upscale target (e.g., 1344×768) and directly drives the 3D upscaler parameters. * **Mode Switch:** Toggle <code>1</code> for Original Image Aspect Ratio, or <code>2</code> for Aspect Ratio Override / Resize.---### ✍️ Subgraph 4: H3 Promptor (Multimodal Engine)* **Primary Role:** Translates and structures creative concepts into MiniMax H3's native multimodal prompt syntax.* **Internal Stages:** 1. **🌐 Translation:** <code>GoogleTranslateTextNode</code> translates non-English inputs to clean English. 2. **👁️ Vision Feature Extraction:** <code>H3_Vision_Analyzer</code> analyzes the initial frame image <code>image_0</code>) to extract character identity, clothing, lighting, and palette. 3. **🧹 VRAM Purge:** Triggers <code>LayerUtility: PurgeVRAM V2</code> to free memory before diffusion starts. 4. **✍️ Multimodal Structuring:** <code>H3_Promptor</code> synthesizes the vision context and creative description into official MiniMax tags: * <code>&lt;d&gt;[Language] ... &lt;/d&gt;</code> * <code>integrated_multimodal_description:</code> * <code>overall_soundscape:</code> * <code>non_diegetic_music:</code> 5. **👁️ Live Preview:** The generated prompt is routed to a <code>ShowText</code> node for immediate inspection.---### 🔍 Subgraph 5: MMH3 2-Pass Latent 3D Upscaler* **Primary Role:** High-resolution spatial and micro-texture refinement.* **Engine:** <code>MMH3UltimateUpscale</code> paired with <code>minimax_h3_latent_upscaler_3d_bf16.safetensors</code>.* **Identity Preservation Setting:** Calibrated to **1 step** with <code>beta57</code> scheduler at **denoise = 0.18**. This guarantees that skin pores, hair strands, and lighting reflections are crisply added without altering facial anatomy or character identity.* **Cinema Sharpening:** Output frames pass through <code>ImageSharpen</code> (Radius: 1, Sigma: 0.8, Strength: 0.35) to remove digital softness and deliver filmic texture.---## 🚀 3. Step-by-Step User Guide### Mode A: Image-to-Video (I2V) — Recommended1. Drag and drop your source image into *<code>1. Start Frame</code>**.2. Set your desired duration in *<code>⏱️ Duration</code>** (default: <code>5.0s</code>).3. In *<code>✍️ H3 Promptor</code>**, type a description of the desired action (e.g. "A woman smiling and turning her head towards the camera on a windy beach").4. Leave *<code>2. End Frame</code>** empty.5. Click **Queue Prompt**.---### Mode B: First &amp; Last Frame (FL2V)1. Load your beginning image into *<code>1. Start Frame</code>**.2. Load your target finishing image into *<code>2. End Frame</code>**.3. In *<code>✍️ H3 Promptor</code>**, set <code>Task Type</code> to <code>FL2VA</code> or describe the transitional trajectory.4. Click **Queue Prompt**. The model will interpolate fluid motion connecting both keyframes.---### Mode C: Reference-to-Video (Ref2V)1. Load your reference character/style image into *<code>3. Reference Image</code>**.2. In **Group 4**, disconnect the latent from <code>MiniMaxH3ImageToVideo</code> and connect the latent output from <code>MiniMaxH3ReferenceToVideo</code> (Node 10) to the Sampler.3. Click **Queue Prompt**.---## ⚙️ 4. Recommended Parameter Reference Table| Parameter | Pass 1 (Base Generation) | Pass 2 (Latent 3D Upscale) | Notes || :--- | :--- | :--- | :--- || **Resolution** | ~864×480 (0.4 MP) | ~1344×768 (0.9 – 1.0 MP) | Driven automatically by Subgraph 3 || **Sampling Steps** | 8 Turbo Steps | 1 Step | Fast 8-step base + 1-step upscale || **Sampler** | <code>euler</code> | <code>er_sde</code> or <code>sa_solver</code> | Ultra-stable temporal coherence || **Scheduler** | <code>beta57</code> | <code>beta57</code> | Calibrated MiniMax curve || **Denoise** | <code>1.00</code> | *<code>0.18 – 0.20</code>** | **Do not exceed 0.25** to avoid face morphing || **Frame Rate** | 24 fps | 24 fps | Standard cinema frame rate || **Sharpening** | None | Radius: 1, Sigma: 0.8, Strength: 0.35 | Sub-second filmic enhancement |---## 💡 5. Best Practices &amp; Pro Tips* **Preserving Facial Likeness:** Keep Pass 2 denoise strictly between <code>0.16</code> and <code>0.20</code>. A denoise of <code>0.50</code> will rewrite facial features and destroy identity.* **Audio Synchronization:** The Audio VAE synthesizes a synchronized soundscape directly from the multimodal prompt. Use headphones to preview ambient audio and spatial sound effects.* **Prompting Secret:** Let <code>H3 Promptor</code> do the heavy lifting. Avoid pasting giant walls of raw tags; instead, describe natural camera movement (*"slow tracking shot"*, "medium close-up") and physical dynamics (*"hair blowing gently in the wind"*).---## 📥 6. Installation &amp; Files1. Place the model checkpoint in: <code>ComfyUI/models/diffusion_models/H3_ErosLite.safetensors</code>2. Place the CLIP model in: <code>ComfyUI/models/clip/qwen3vl_32b_heretic_minimax_h3_nvfp4.safetensors</code>3. Place the VAEs in: <code>ComfyUI/models/vae/minimax_h3_video_vae_fp16.safetensors</code> <code>ComfyUI/models/vae/minimax_h3_audio_vae_fp32.safetensors</code>4. Drag and drop the workflow file <code>H3_ErosLite_2Pass_Clean_Workflow.json</code> into ComfyUI.

Files with SHA256: 943052dd16187e235c647a36d78fc08cfe28111e7280bd9d4b69390f538f873e