AI Video Generation API: Make Finished Videos from Your Own Code

Two kinds of AI video generation API

Before you pick an API, know which output you need. They solve different problems.

If you are building a content pipeline (a blog that gets a video for every post, a product feed turned into ads, an agent that runs a channel), a finished-video API saves you the voice, editing and captioning layers. More on that in AI content pipelines with video.

What the Autoclips API covers

The API does what the app does. The API page lists 258 operations, each with an example you can copy, grouped by product:

• Autoclips V3: cinematic narrated story videos, quick or step by step (script, cast, scenes, render).
• Character Explainer: cartoon characters explaining a topic over gameplay.
• 3D Animated Studio: 3D animated shorts, including an autopilot with a spending limit.
• Stickman videos: explainers, skits and stories, story plans, series and automations.
• Shared: account and credits, quotes, jobs, media uploads, voices, timeline edits, covers, outros and publishing to connected social accounts.

Your first calls

1. 1

   Create a key with only the permissions it needs

   In Autoclips, open Settings → API and create a key. Pick its permissions: read (see projects and videos), edit (free changes), spend (anything that costs credits), delete, and publish (post to your social accounts). Permissions don't include each other. The key is shown once; put it in an environment variable.

2. 2

   Check the key

   Every address starts with https://app.autoclips.app/api/v1. Send the key as a Bearer token:

   curl https://app.autoclips.app/api/v1/me -H "Authorization: Bearer $AUTOCLIPS_API_KEY"

   It answers with your account and what the key may do. Then read the valid choices for your product with GET /api/v1/options, so you only send values that exist.

3. 3

   Get a quote, then start the video

   Anything that spends credits needs your approval. Either send max_credits (the most you accept), or ask POST /api/v1/quotes with the operation and its exact input and get a quote_token that works once, for 10 minutes. If you send neither, nothing runs and you get a 402 confirmation_required with the quote.

   Requests that create something or spend credits also need an Idempotency-Key header, so a retry never makes a second video.

4. 4

   Check the job until it finishes

   Making a video answers 202 with a job. Ask GET /api/v1/jobs/{jobId} again after the suggested wait until the status is succeeded (with the video URL), failed or canceled (each says why).

Built so code can't overspend

• Prices are worked out on the server, never taken from the request, and a quote is an upper bound: you are never charged more than it.
• Daily credit caps per key and per account stop a runaway loop. Hitting one returns 402 daily_credit_cap_reached, and nothing runs.
• Scoped keys: a key without spend can't spend; one without publish can't post.
• Stable error codes (insufficient_credits, quote_exceeds_max_credits, insufficient_scope and more), each with a docs_url.
• Rate-limit headers on every answer, and Retry-After on every 429.

The same API for AI apps (MCP)

The operations are also an MCP server at https://app.autoclips.app/api/mcp, so Claude, ChatGPT, Cursor and other AI apps can use them after a sign-in, no key needed. That is the fastest way to try the API before writing code: connect it and ask for a video in plain words. See how to make videos with Claude and the MCP page.

What people build with it

• Blog to video: every new post in your CMS becomes a 60 second video for Shorts and TikTok.
• No-code flows: Zapier, Make or n8n call the API as plain HTTP requests.
• Channel agents: an AI agent picks topics, makes videos and posts them under a daily cap. See automating a YouTube channel with an AI agent.
• Apps for clients: agencies make videos for many brands from one system.