AI generation | Simplified Developer Docs

1. Define the output

Describe the subject, audience, style, dimensions or aspect ratio, and intended channel. Read the selected brand kit before generating branded content.

Create a square product-launch image for Acme. Use the brand palette, leave clear space for a headline, and save the finished image to our Launch project.

2. Discover a model and its inputs

For images, use api_listImageModels, api_getImageModel, and api_getImageModelFields. For video, use api_listVideoModels, api_getVideoModel, and api_getVideoModelFields. api_getModelFields provides additional model-field discovery where applicable.

Choose a model that supports the requested operation. Supported sizes, durations, reference inputs, and generation settings vary by model. Read the schema before supplying them; do not reuse another model’s settings without checking.

Check api_getCreditBalance before a large generation request. Credit balance alone does not establish the exact cost of a job.

3. Generate an image or video

Call api_generateImage or api_generateVideo with the discovered inputs. Use uploaded asset references or URLs in the format the selected model requires.

Some calls wait for a completed result; others can return a job or variation that still needs checking. Keep all returned identifiers. Follow Jobs and results if the result is still processing or your client times out.

For video variations, use api_getVideoVariation with the returned video and variation identifiers.

4. Create a voiceover

Use api_listVoices and api_getVoice to choose an available voice. Submit the script and supported settings with api_generateAudio. Use the returned job information if processing continues after the initial call.

Generate a voiceover for this 30-second script using an available English voice. Show me the chosen voice before generating it.

5. Save and reuse the output

Inspect the result before accepting it. If it is a temporary URL, import it as an asset. Add its asset ID to the intended project item or social draft. Report the saved resource IDs and the actual processing state, not just the submitted prompt.