AI PPT

Create and read AI PPT results.

AI PPT API

Use slides to create an editable presentation with PPT and PDF results.

Create

POST /creations
{  "type": "slides",  "prompt": "Create an onboarding presentation that explains MCP to new product managers",  "language": "en",  "aspect_ratio": "16:9",  "scene_count": 8}

Choose page count, ratio, quality, and style from /catalog. AI PPT does not use voice_id or duration_seconds.

Manage

The priority is usually not playback. It is:

  • page count
  • cover image
  • per-page assets
  • document exports

So the detail response usually matters for:

  • artifacts[]
  • scenes[]
  • main_url

Revise

Public OpenAPI does not currently expose scene-level editing for AI PPT. To revise the presentation, create a new result with the updated prompt and settings.

Export

The most common exports are:

  • PPT
  • PDF

Use the export assets returned by the detail response.

Practical Guidance

  • treat slides as a document workflow, not a video workflow
  • do not send voice_id or duration_seconds
  • page, ratio, and quality constraints should all come from /catalog

Choose AI PPT for a reusable presentation

Use AI PPT when the primary deliverable is a deck that someone will read, present, export, or adapt. State the audience, the decision or learning objective, and the intended sequence of pages. An onboarding deck should identify the role, the first decisions to make, and the desired practical outcome rather than only requesting “an onboarding presentation.”

Work from document constraints

Read /catalog and use its supported page count, ratio, quality, and style values. scene_count shapes the presentation scope; voice_id and duration_seconds do not belong to an AI PPT request. Create the job, poll it to completion, and use the detail response as the authority for main_url, scenes[], and the PPT or PDF assets returned in artifacts[].

Public OpenAPI does not currently offer scene-level editing for this type. When the deck needs a different scope, structure, or instruction, create a revised result with a more precise prompt instead of attempting a video edit action.