<div data-youtube-video="https://youtu.be/R3W2ab1_TzY"><iframe src="https://www.youtube.com/embed/R3W2ab1_TzY" width="853" height="480" allowfullscreen></iframe></div> 
 
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2098775143422537729?inviteCode=rh-v1111Bilibili tutorial video: https://www.bilibili.com/video/BV1vFYz6HECU/# Sigma two-pass — 4K-target post-processing## Purpose and preparationResolve the matching custom-node/runtime versions and assets. Load authorized media and inspect active inputs, actual schedules, effective dimensions and every save destination before a small test.Only image 139 is active. Image256 remains wired but bypassed; images 257–260 are bypassed/disconnected. The generation selector requests 16:9,0.5MP,multiple32, overriding cached dimensions. Duration 5 requests 124 frames arithmetically, not a measured five-second output.The active Ref2V LoRA is named 8step, approximately strength 1. BasicScheduler stores simple/10/denoise1, followed by H3SigmaRefiner(extra_steps 1,start 0.7,end 0,cosine), then SplitSigmas at 6. Because refinement precedes the split,6+4 actual steps are not established. The model sigma shift is video 12/audio 3; Euler, BasicGuider and fixed-control noise are shared.First DENOISED_OUTPUT is separated. Video enters a learned FP16 latent upscaler configured multiplier 2/align 32/CUDA/chunking; audio bypasses that upscaler and is concatenated into the second latent. Second OUTPUT feeds video/audio decoding. This is not proof of unchanged audio, exact raster or quality improvement. Native generated audio, not source soundtrack or conditioner mux_audio, reaches assembly.Decoded images pass through SeedVR2, linked target2160, then a resize request3840×2160, Lanczos, keep_proportion=resize, divisible_by2, CPU. Audio bypasses spatial processing. These are post-decoding targets, not native generation dimensions or measured results. Cached resize diagnostics are not render evidence.Interpolation requests 2x,motion auto,threshold0.24,keep_audio on. This is temporal, not spatial scaling. For generation presets,24→48fps is configured intent, not measured cadence/frame count/duration. Combined VIDEO and explicit images/audio are connected; precedence requires implementation review. VIDEO output feeds SaveVideo with auto format/codec.## Files and selected assets- [Workflow JSON]()- [Selected cover]()- `qwen3vl_32b_minimax_h3_int8_convrot.safetensors`- `minimax_h3_video_vae_fp16.safetensors`- `minimax_h3_audio_vae_fp32.safetensors`- `minimax_h3_hybrid_fl2va_ref2va_b25-49-int8.safetensors`- `minimax_h3_ref2v_turbo_8step_v1.0_768p_comfyui_bf16.safetensors`- `seedvr2_ema_3b_fp8_e4m3fn.safetensors`- `minimax_h3_latent_upscaler_3d_fp16.safetensors`- `taeh3.safetensors`## Version, evidence and rightsThis is a local descriptive draft, not a published upgrade. Files01–08 are byte-identical to originals; only the standalone ninth file's historical video-preview field is cleared in its new copy. Originals remain unchanged.No model inference, visual/audio acceptance, wall-clock benchmark or license verification was performed. Covers are selected assets, not accepted render evidence. Weights and external input media are not bundled. Check workflow, nodes, models, LoRAs, source media and cover rights separately; no blanket commercial license is asserted.[Source-provided tutorial](https://www.youtube.com/watch?v=R3W2ab1_TzY) · [Source-provided workflow entry](https://www.runninghub.cn/post/2098400619433381889/?inviteCode=rh-v1111)Links are not live-verified. The workflow is supplied through AI-KSK.

MiniMax H3 Sigma Enhanced Dual-Pass Universal Video Generation 4K

<div data-youtube-video="https://youtu.be/R3W2ab1_TzY"></div> 
 
RunningHub workflow page: https://www.runninghub.ai/zh-cn/post/2098775143422537729?inviteCode=rh-v1111Bilibili tutorial video: https://www.bilibili.com/video/BV1vFYz6HECU/# Sigma two-pass — 4K-target post-processing## Purpose and preparationResolve the matching custom-node/runtime versions and assets. Load authorized media and inspect active inputs, actual schedules, effective dimensions and every save destination before a small test.Only image 139 is active. Image256 remains wired but bypassed; images 257–260 are bypassed/disconnected. The generation selector requests 16:9,0.5MP,multiple32, overriding cached dimensions. Duration 5 requests 124 frames arithmetically, not a measured five-second output.The active Ref2V LoRA is named 8step, approximately strength 1. BasicScheduler stores simple/10/denoise1, followed by H3SigmaRefiner(extra_steps 1,start 0.7,end 0,cosine), then SplitSigmas at 6. Because refinement precedes the split,6+4 actual steps are not established. The model sigma shift is video 12/audio 3; Euler, BasicGuider and fixed-control noise are shared.First DENOISED_OUTPUT is separated. Video enters a learned FP16 latent upscaler configured multiplier 2/align 32/CUDA/chunking; audio bypasses that upscaler and is concatenated into the second latent. Second OUTPUT feeds video/audio decoding. This is not proof of unchanged audio, exact raster or quality improvement. Native generated audio, not source soundtrack or conditioner mux_audio, reaches assembly.Decoded images pass through SeedVR2, linked target2160, then a resize request3840×2160, Lanczos, keep_proportion=resize, divisible_by2, CPU. Audio bypasses spatial processing. These are post-decoding targets, not native generation dimensions or measured results. Cached resize diagnostics are not render evidence.Interpolation requests 2x,motion auto,threshold0.24,keep_audio on. This is temporal, not spatial scaling. For generation presets,24→48fps is configured intent, not measured cadence/frame count/duration. Combined VIDEO and explicit images/audio are connected; precedence requires implementation review. VIDEO output feeds SaveVideo with auto format/codec.## Files and selected assets- [Workflow JSON]()- [Selected cover]()- `qwen3vl_32b_minimax_h3_int8_convrot.safetensors`- `minimax_h3_video_vae_fp16.safetensors`- `minimax_h3_audio_vae_fp32.safetensors`- `minimax_h3_hybrid_fl2va_ref2va_b25-49-int8.safetensors`- `minimax_h3_ref2v_turbo_8step_v1.0_768p_comfyui_bf16.safetensors`- `seedvr2_ema_3b_fp8_e4m3fn.safetensors`- `minimax_h3_latent_upscaler_3d_fp16.safetensors`- `taeh3.safetensors`## Version, evidence and rightsThis is a local descriptive draft, not a published upgrade. Files01–08 are byte-identical to originals; only the standalone ninth file's historical video-preview field is cleared in its new copy. Originals remain unchanged.No model inference, visual/audio acceptance, wall-clock benchmark or license verification was performed. Covers are selected assets, not accepted render evidence. Weights and external input media are not bundled. Check workflow, nodes, models, LoRAs, source media and cover rights separately; no blanket commercial license is asserted.[Source-provided tutorial](https://www.youtube.com/watch?v=R3W2ab1_TzY) · [Source-provided workflow entry](https://www.runninghub.cn/post/2098400619433381889/?inviteCode=rh-v1111)Links are not live-verified. The workflow is supplied through AI-KSK.

minimaxH3SigmaEnhanced_v10.json

v1.0

<div data-youtube-video="https://youtu.be/R3W2ab1_TzY"><iframe src="https://www.youtube.com/embed/R3W2ab1_TzY" width="853" height="480" allowfullscreen></iframe></div> 
 
AI-KSK ComfyUI workflow: MiniMax H3 Sigma Dual-Pass 4K. Maximum-fidelity tier of the pack: dual-pass Sigma reinforcement sampling, 3D latent upscaling, and SeedVR2 post upscaling to UHD 3840x2160. 
 
Main features: 
- MiniMax H3 hybrid FL2VA/Ref2VA INT8 UNET with Qwen3-VL 32B text encoder, video and audio VAEs for native audio generation. 
- Dual-pass Sigma reinforcement: 0.5 MP first pass followed by a second pass with the 768p Ref2V Turbo 8-step LoRA at full strength and 10-step Euler sampling. 
- MiniMax H3 3D latent upscaler (FP16, 2x multiplier) between passes for higher internal detail. 
- SeedVR2 3B FP8 video upscaling, normalized to UHD 3840x2160. 
- Native audio preserved, DLSS 2x frame interpolation (24 to 48 FPS). 
- The longest render in the 9-workflow pack; reserve it for final cuts and hero shots. 
 
Suggested workflow: 
1. Import the JSON into ComfyUI and install the model files and custom nodes referenced by the workflow. 
2. Provide the required text, image, or video inputs for the selected path. 
3. Adjust resolution, steps, denoise, and prompt controls, then run the connected SaveVideo output. 
 
 
RunningHub workflow: 
https://www.runninghub.ai/zh-cn/post/2098775143422537729?inviteCode=rh-v1111 
 
AI-KSK RunningHub workflow hub: 
https://www.runninghub.ai/user-center/1892566146468511746?inviteCode=rh-v1111 
 
Bilibili tutorial: 
https://www.bilibili.com/video/BV1vFYz6HECU/ 
 
RunningHub registration and AI tools: 
https://www.runninghub.ai/?inviteCode=rh-v1111 
https://www.runninghub.ai/ai-generator/2046514150500524034?inviteCode=rh-v1111 
https://www.runninghub.ai/vip-rights/3?inviteCode=rh-v1111 
https://rhtv.runninghub.ai/projects?inviteCode=rh-v1111 
 
Stable global access: 
https://lelians.com/#/register?code=mFIJ5tZn 
 
TokenRhythm AI model credits: 
https://tokenrhythm.studio/i/rf_tr_kM6wekDquDDZQD4PUNmgiuHA 
 
Youyun Cloud GPU: 
https://passport.compshare.cn/register?referral_code=A7fTQUYY9lCDFJwA6MMvoq 
https://www.notion.so/2b72459a3ac6808f8282f2c687ed8f52 
 
RunPod Cloud GPU: 
https://runpod.io/?ref=sxk5n5m7 
 
Support AI-KSK: 
https://ko-fi.com/aiksk

MiniMax H3 Sigma Dual-Pass 4K

<div data-youtube-video="https://youtu.be/R3W2ab1_TzY"></div> 
 
AI-KSK ComfyUI workflow: MiniMax H3 Sigma Dual-Pass 4K. Maximum-fidelity tier of the pack: dual-pass Sigma reinforcement sampling, 3D latent upscaling, and SeedVR2 post upscaling to UHD 3840x2160. 
 
Main features: 
- MiniMax H3 hybrid FL2VA/Ref2VA INT8 UNET with Qwen3-VL 32B text encoder, video and audio VAEs for native audio generation. 
- Dual-pass Sigma reinforcement: 0.5 MP first pass followed by a second pass with the 768p Ref2V Turbo 8-step LoRA at full strength and 10-step Euler sampling. 
- MiniMax H3 3D latent upscaler (FP16, 2x multiplier) between passes for higher internal detail. 
- SeedVR2 3B FP8 video upscaling, normalized to UHD 3840x2160. 
- Native audio preserved, DLSS 2x frame interpolation (24 to 48 FPS). 
- The longest render in the 9-workflow pack; reserve it for final cuts and hero shots. 
 
Suggested workflow: 
1. Import the JSON into ComfyUI and install the model files and custom nodes referenced by the workflow. 
2. Provide the required text, image, or video inputs for the selected path. 
3. Adjust resolution, steps, denoise, and prompt controls, then run the connected SaveVideo output. 
 
 
RunningHub workflow: 
https://www.runninghub.ai/zh-cn/post/2098775143422537729?inviteCode=rh-v1111 
 
AI-KSK RunningHub workflow hub: 
https://www.runninghub.ai/user-center/1892566146468511746?inviteCode=rh-v1111 
 
Bilibili tutorial: 
https://www.bilibili.com/video/BV1vFYz6HECU/ 
 
RunningHub registration and AI tools: 
https://www.runninghub.ai/?inviteCode=rh-v1111 
https://www.runninghub.ai/ai-generator/2046514150500524034?inviteCode=rh-v1111 
https://www.runninghub.ai/vip-rights/3?inviteCode=rh-v1111 
https://rhtv.runninghub.ai/projects?inviteCode=rh-v1111 
 
Stable global access: 
https://lelians.com/#/register?code=mFIJ5tZn 
 
TokenRhythm AI model credits: 
https://tokenrhythm.studio/i/rf_tr_kM6wekDquDDZQD4PUNmgiuHA 
 
Youyun Cloud GPU: 
https://passport.compshare.cn/register?referral_code=A7fTQUYY9lCDFJwA6MMvoq 
https://www.notion.so/2b72459a3ac6808f8282f2c687ed8f52 
 
RunPod Cloud GPU: 
https://runpod.io/?ref=sxk5n5m7 
 
Support AI-KSK: 
https://ko-fi.com/aiksk

Files with SHA256: 06f0e62dd7168a98fdb0ef762358d1780a19f730249e0bd56b0811ed8e6dbaf9