Generate AI images, video and audio from a prompt

[Jesus-Shotstack]

Generate AI images, video and audio from a prompt

Generating an asset from n8n today needs a raw HTTP Request node and a hand-built polling loop. The node now does it in one step, and an AI agent can list models and their options before spending credits.

• Generate Asset waits by default, up to Give Up After (5 minutes, max 10): n8n runs items one at a time, so six items at 10 minutes hit its one-hour execution limit.
• Only a 200 ends the wait. A 202 saying done can still lack its URL.
• No Idempotency Key: identical requests already share one cached result, billed once. A field can be added later; removing one breaks saved workflows.
• Render Asset's Edit now shares the Model Options parser, so {} fails as empty before it reaches the API.
• @shotstack/schemas 1.22.0 carries the generation endpoints the spec checks compare against.

Ships enabled: n8n has no feature flags, so the resource reaches users only with a release.

Verify: node scripts/review-check.mjs passes 40/40 with 86 tests. test/generation-guards.mjs drives the wait through 202, 401, 404, dropped polls and the six-item limit.
Risk: generation bills credits in Sandbox too. Rollback is a revert; no released version has this resource.