<h1 id="minimax-h3-multiangle-workflow-for-comfyui">MiniMax H3 Multiangle Workflow for ComfyUI</h1>Change the camera viewpoint of a reference image with MiniMax H3. Start with a single image and explore a three-quarter view, a high or low angle, or a tighter crop. The interactive 3D camera panel lets you adjust the horizontal angle, vertical angle, and framing before generating the edit.Use Match Background View to control how MiniMax H3 handles the scene. Turn it on to request the original surroundings from the new viewpoint. Turn it off to place the subject against a plain background. You can also use Additional Prompt to ask the model to preserve details such as the outfit, hairstyle, or lighting.The workflow translates camera settings into editing instructions. The requested angle guides the generation, so the exact viewpoint and visible details can vary between outputs.<h2 id="getting-started">Getting started</h2><ol><li>Use ComfyUI 0.44.0 or later and install both custom nodes listed below.</li><li>Download the four model files and place them in the corresponding folders.</li><li>Open <code>P3_image_MultiangleCamera.json</code> in ComfyUI and select a reference image in Load Image.</li><li>Adjust the camera and background settings, then queue the workflow. The included Save Animated WEBP node saves the result to ComfyUI’s standard output folder.</li></ol><h2 id="full-setup-guide"><a target="_blank" rel="ugc" href="https://github.com/ukr8b3g-cmyk/ComfyUI-H3-Camera-Prompt-Adapter#readme">Full setup guide</a></h2><h2 id="left-and-right-camera-angles">Left and right camera angles</h2>Left and right are relative to the camera’s starting viewpoint. MiniMax H3 can sometimes favor the original angle or show the subject from the opposite side. If the result isn’t what you intended, try a few nearby horizontal angles, compare both sides, and experiment with a different seed.<h3 id="camera-angle-and-background-limitations">Camera angle and background limitations</h3>In my testing, the MiniMax H3 workflow handled a narrower range of camera angles reliably than the Qwen-based workflow. Some settings may produce a view close to the original or show the subject from the opposite side. If an angle doesn’t work as expected, try nearby horizontal or vertical settings.Match Background View asks the model to show the original scene from the new camera position. The background may stay close to the original view even when the subject’s angle changes, or the model may rearrange objects and alter parts of the scene. Viewpoint accuracy and background consistency can vary considerably across angles and reference images.<h2 id="required-custom-nodes">Required custom nodes</h2><ul><li>Qwen Multiangle Camera provides the interactive 3D camera controls. <a target="_blank" rel="ugc" href="https://github.com/jtydhr88/ComfyUI-qwenmultiangle">GitHub</a></li><li>H3 Camera Prompt Adapter converts the camera prompt and position into editing instructions for MiniMax H3. <a target="_blank" rel="ugc" href="https://github.com/ukr8b3g-cmyk/ComfyUI-H3-Camera-Prompt-Adapter">GitHub</a> · <a target="_blank" rel="ugc" href="https://registry.comfy.org/nodes/comfyui-h3-camera-prompt-adapter">Comfy Registry</a></li></ul><h2 id="model-downloads">Model downloads</h2>ModelInstallation folderFileMiniMax H3 FL2VA diffusion model<code>ComfyUI/models/diffusion_models/</code><a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/MiniMax-H3/blob/main/diffusion_models/minimax_h3_fl2va_pruned_int8_convrot.safetensors">minimax_h3_fl2va_pruned_int8_convrot.safetensors</a>Qwen3-VL 32B NVFP4/AWQ text encoder<code>ComfyUI/models/text_encoders/</code><a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/MiniMax-H3/blob/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors">qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors</a>P-H3-Image VAE<code>ComfyUI/models/vae/</code><a target="_blank" rel="ugc" href="https://huggingface.co/PrunaAI/P-H3-Image-Comfy/blob/main/vae/p-h3-image_vae_int8_convrot.safetensors">p-h3-image_vae_int8_convrot.safetensors</a>P-H3-Image LoRA<code>ComfyUI/models/loras/</code><a target="_blank" rel="ugc" href="https://huggingface.co/PrunaAI/P-H3-Image-Comfy/blob/main/loras/p-h3-image_lora_bf16.safetensors">p-h3-image_lora_bf16.safetensors</a>

Files with SHA256: c73d79543174c90bbfc8c1375e273376c77adb08c38c6dd02a58a841634c18b5