Project IMAGINE
Intelligent Multimodal AI for Generative Ideation, Narrative & Exploration.
IMAGINE is developed by eXtended Reality Lab (XR-Lab), with generous support from the Gensler AI Grant and additional funding and resources from University of Cincinnati.
The goal of IMAGINE is to create a unified, multimodal Generative AI platform that supports UC faculty and student research, teaching, experimentation, and creative work. The platform integrates multiple open-source AI models hosted locally on high-performance computers at XR-Lab, equipped with powerful GPUs and large.
Access is currently available only within the UC network, or VPN. Please click the link above to access the IMAGINE platform. Once you log in, you will see the IMAGINE interface, which provides access to various AI models for image, 3D, and video generation.
Image generation is built on Stable Diffusion models.
We currently support two workflows:
Text-to-Image — Generate images directly from text prompts.
Image-to-Image — Use an existing image together with a text prompt to generate new visual content.
For better results, you may use your preferred chat-based AI/LLM, such as ChatGPT or Claude, to help develop and refine your prompt and negative prompt before submitting a job.
You can also adjust parameters such as image resolution and CFG values to control the generation process.
Stable Diffusion WebUI (Advanced Features)
The IMAGINE portal only offers basic image generation with limited controls. For more advanced capabilities, you can access the full Stable Diffusion WebUI, which includes mask-based inpainting, ControlNet, upscaling, and various LoRAs for creating 360° panoramas, Midjourney-style images, prompt travel animations, and more.
To access the full SD WebUI, enter the same password used for the IMAGINE portal. Once authenticated, you will find a link to the full WebUI.
Tutorial: AI for imaging generation.
3D generation is built on the Hunyuan 3D model and currently supports:
Single-View to 3D — Generate a 3D model from a single image.
Multi-View to 3D — Generate a 3D model using multiple reference images.
Please note that the Multi-View to 3D server is available only after 7:00 PM, when additional GPU resources become available.
Video generation is built on the open-source Wan 2.2 model. Multiple NVIDIA RTX GPU workstations at XR-Lab are connected to the platform to support video generation.
Additional rendering machines become available after 7:00 PM, so computationally intensive jobs may run more efficiently during evening hours.
To maintain fair access to computing resources, each video-generation job has a one-hour rendering limit. The system will automatically terminate any job that remains unfinished after one hour.
Please keep this limit in mind when increasing parameters such as resolution, number of frames, or sampling steps, as higher settings can significantly increase rendering time.
Recommended Image-to-Video Settings
| Settings | Test | Final |
|---|---|---|
| Resolution | 960 × 544 | 1280 × 720 |
| Frames / FPS | 81 / 16 | 81 / 16 |
| Steps | 25 | 30 |
| CFG | 4 | 4 |
Post-processing: Upscale 2× and apply frame interpolation.
Prompt: Slow cinematic camera orbit. Smooth, steady movement. Architecture remains stable and unchanged. Subtle atmospheric movement and consistent lighting
Negative Prompt: warping, morphing, distorted architecture, deformed buildings, unstable geometry, changing building shapes, flickering, jitter, camera shake, sudden movement, rotation, blurry, artifacts
For Scenes with Human Faces
Add to Prompt: Slow cinematic camera push forward. Smooth, steady movement. Architecture remains stable and unchanged. Subtle atmospheric movement and consistent lighting. same person, consistent facial features, detailed natural face, sharp eyes, stable facial identity, realistic skin
Add to Negative Prompt: warping, morphing, distorted architecture, deformed buildings, unstable geometry, changing building shapes, flickering, jitter, camera shake, sudden movement, rotation, blurry, artifacts, deformed face, distorted face, changing face, blurry face, asymmetrical eyes, malformed eyes, duplicate face, facial artifacts, morphing
Tutorials
Questions and Feedback
IMAGINE is an evolving research and creative platform. We welcome your feedback as we continue to test new models, improve workflows, and expand its capabilities.
Please contact Ming Tang if you have any questions, encounter technical issues, or have suggestions for the platform.
XR-Lab AI Server Usage Policy
AI Rendering and Content Generation
The XR-Lab AI Server is a shared resource supporting academic research, teaching, and creative exploration. All users are expected to follow these guidelines to ensure fair access, responsible use, and efficient performance.
- Authorized Use: The server is intended for UC-related academic, research, and creative projects, including AI image, video, and 3D content generation. Commercial or personal use requires prior approval.
- Fair Resource Sharing: GPU resources are shared among users. Avoid running unnecessary jobs or occupying multiple GPUs simultaneously. Large rendering tasks should be scheduled during off-peak hours, preferably overnight.
- Job Management: Test prompts and settings at lower resolutions before submitting high-resolution or lengthy rendering jobs. Remove unnecessary files and completed jobs to conserve storage.
- Responsible AI Use: Users must respect copyright, intellectual property, privacy, and university policies. Do not upload confidential, sensitive, or restricted materials without authorization. Inappropriate or unlawful content generation is prohibited.
- Data and Storage: Users are responsible for backing up their generated content. Temporary files and inactive projects may be deleted periodically to maintain server performance.
- Academic Integrity: AI-generated content should be appropriately acknowledged in academic submissions and publications, following course and university guidelines.
- Technical Support: Report technical problems to XR-Lab staff.
Important: Access to the AI server is a privilege. Repeated misuse or excessive resource consumption may result in restricted access.
XR-Lab | University of Cincinnati | Digital Futures
