letz-ai/letzai-skill
Overview
This skill integrates LetzAI API for generating and editing AI images and videos, plus upscaling and using custom trained models via @modelname. It supports Nano Banana Pro, Flux2 Max, SeeDream for images and VEO/Kling for videos, and is built for content creation apps and automation. Use it to automate asset pipelines, generate creative media, or incorporate personalized trained models into production workflows.
How this skill works
The skill issues authenticated POST requests to LetzAI endpoints (/images, /videos, /image-edits, /upscales) and then polls the corresponding GET endpoints until jobs reach status "ready." Prompts may include @modelname to invoke custom trained models. Responses provide direct URLs for generated images, edited outputs, upscaled assets, or videoPaths once processing completes.
When to use it
- Automating image/video production for social, ads, or e-commerce content
- Generating images with custom-trained person, object, or style models via @modelname
- Performing context-based image edits (replace background, change objects)
- Creating short videos from a source image and AI motion models
- Upscaling generated or existing images for higher-resolution delivery
Best practices
- Choose baseModel based on quality vs cost (Nano Banana Pro for highest quality, Flux2 for cost balance)
- Include @modelname only after confirming trained model exists with GET /models
- Respect asynchronous workflow: poll GET endpoints at recommended intervals and implement exponential backoff for 429s
- Limit image edits to a maximum of 9 reference images and validate aspect_ratio and dimensions before request
- Handle errors robustly: check HTTP codes (401, 402, 400, 429, 500) and surface human-readable messages to users
Example use cases
- A mobile app pipeline that generates product photos using an object model @my_product and then upscales to 4K for marketing
- An influencer tool that edits selfies with context editing to change backgrounds and apply a trained style @vintage_style
- Automated ad creative generation: create image variants, then convert a selected image into a short VEO or Kling video
- A studio workflow that batches image generation, polls status, and stores imageVersions.original URLs in a CDN
- Quality assurance script that lists available trained models, picks the right class, and tags prompts with @modelname
FAQ
Send a Bearer token in the Authorization header and set Content-Type: application/json. Get an API key via letz.ai/subscription.
Can I train models via the API?
No. Model training is done in the LetzAI web interface; the API can list and use trained models via GET /models and the @modelname prompt syntax.
