GlanceAI Model Disclosure
AI models and capabilities
This public page describes Glance cloud Widget capabilities, underlying models and processing services. No account, download or purchase is required to read it.
Updated: October 22, 2026
Which AI capabilities does Glance offer?
| Capability | Current scope |
|---|---|
| Text to image | No current Widget generates an image from text alone. |
| Image to image and image editing | Requires an input image. Includes regeneration, upscaling, outfit changes, outpainting, watermark removal and background removal. Regeneration, outfit changes and outpainting create new image content. |
| Image to video | No current Widget generates a video from a still image. |
| Video inpainting | Video Watermark Removal processes an existing video and returns an edited video. It does not generate video from a still image. |
| Image understanding | Image to Prompt describes an input image in text; it does not generate images. |
Models by Widget
The table lists the models used by the current public Widget catalog. Version information requiring further confirmation is noted separately. RunningHub is a processing platform, not an underlying model name. Official references identify the models and do not imply endorsement of Glance.
| Widget | What it does | Underlying model | Runs on |
|---|---|---|---|
| Upscale | Image to image: upscales and re-renders detail, up to 4K | Google Gemini 3.1 Flash Image (Nano Banana 2) | RunningHub |
| Image Watermark Removal | Image inpainting: removes marks and text and fills the area | Black Forest Labs FLUX.1 Kontext [dev] (Nunchaku INT4 build) | RunningHub |
| Portrait Background Removal | Segmentation: returns a transparent-background image | BRIA RMBG-1.4 + ViTMatte (ComfyUI LayerStyle RmBgUltra V2) | RunningHub |
| Image to Prompt | Image understanding: writes a text description | Google Gemini Used for image understanding and text description; the exact model and version remain to be verified. |
RunningHub |
| Video Watermark Removal | Video inpainting: cleans 8-second segments and rejoins them | MiniMax-Remover (based on Wan2.1-1.3B) + Ultralytics YOLO watermark detector | RunningHub |
| Redraw & Reframe | Regenerates an input image at a selected aspect ratio | Krea 2 | RunningHub |
| Outfit Swap | Generates an outfit edit from a person photo and clothing reference | Qwen-Image-2.1 | RunningHub |
| Expand Image | Generates content beyond the input image boundaries | Black Forest Labs FLUX.2 | RunningHub |
| Image Hosting | Uploads and returns a share link; no AI involved | — | Freeimage.host (images from mainland China) / Cloudflare R2 |
Image to Prompt version note: this feature uses Google Gemini to analyze an input image and return text, not to generate images. The model family is confirmed; the exact model and version remain to be verified.
Media processing and use
When you run a cloud AI Widget, the selected media is uploaded to Glance and processed through RunningHub. Image Hosting stores and shares files without using an AI model. Glance does not use your media to train models.
Upload only content you own or have permission to process. Obtain the subject’s consent before using their photo for an outfit change. See our Privacy Policy, Terms of Service and Acceptable Use Policy. See Pricing for prices.