Complete workflows for model training, image generation, upscaling, public-model refinement, and video generation.
End-to-End Examples
These guides are organized around reusable setup first, then concrete use cases. Start by creating and training a model once for the person, creator, founder, or staff member you want to generate. After the model is ready, choose a use-case recipe and follow its asset pipeline.
Training polling
Model training normally takes 20 to 25 minutes. Poll GET /models/{modelID} every 5 minutes while
training is running. Avoid tight polling loops.
How the guides fit together
| Page | Use it for |
|---|---|
| Setup and Training | API helpers, model creation, training-image upload, training, and polling. |
| UGC Product Video | Creator-style product stills, upscaled stills, polished ads, and video. |
| Local Business Campaign | Founder or staff assets for ads, landing pages, posters, and short clips. |
| Personal Website Hero Assets | Professional portraits, website banners, and intro videos. |
The use-case pages assume you have completed setup and training and have a ready modelId. Each
recipe then follows the same production shape:
- Generate an identity-preserving image with the trained model.
- Upscale the selected generated image.
- Improve or recompose the image with a public image model like GPT Image 2.
- Use the final still as the first frame for video generation.
Guides
Setup and Training
Create a model, upload training images, trigger training, and poll every 5 minutes.
UGC Product Video
Build a creator product ad from trained identity image to final video.
Local Business Campaign
Create campaign stills and video for local business marketing.
Personal Website Hero Assets
Generate portraits, banners, and a subtle intro video for a personal site.