Contents
Advances in generative AI make it possible to automate the creation of news videos. An AI voice can read a summary of a story like a news presenter, while the video shows the headline and captions. Add to this the wide availability of news APIs and the Shotstack video editing API, and you can create personalized news videos at scale.
Video is now one of the main ways people get their news. The Reuters Institute Digital News Report 2026 found that 77% of people globally consume online news video each week.
In this guide, we will show you how to generate news videos automatically using AI and a few APIs. We will use the World News API to get the content of a news article and OpenAI’s Responses API to summarize it. Then the Shotstack Edit API turns the summary into a narrated video with captions, in a single render.
TL;DR: Fetch a story from a news API, ask an LLM for a 25-second summary, and put that summary in the prompt of an audio asset in a Shotstack edit. One render request generates the newsreader voice and a background image, captions the voice word by word and adds the headline.
About the APIs in this guide
About Shotstack
Shotstack is a cloud-based video automation platform that includes APIs, design and workflow tools to create and edit videos at scale. In this guide, we use the Edit API to generate the newsreader voice and the background image and to edit everything into a video, all in one request.
About the World News API
The World News API provides news articles from around the world. Unlike many news APIs, it includes the complete body of the article, not just the article URL. It also includes a headline, image, search functionality and is easy to use. Its free plan searches the last month of news.
About the OpenAI Responses API
The OpenAI Responses API gives you the models behind ChatGPT in your own code. It can summarize a body of text into a shorter, more concise version. We’ll use it to write the script for the voice over.
Prerequisites
To follow along with this guide, you will need:
- A Shotstack account and API key
- A World News API key
- An OpenAI API key
- Basic knowledge of interacting with APIs and cURL
Rendering and AI generation in Shotstack use credits. See pricing for the rates.
Generate an AI news video
The following steps show you how to call each API to generate a news video. We use cURL in this guide, and the last section shows how the Shotstack step looks in code, so you can automate it inside an application or workflow.
Fetch news from the World News API
The first step is to call the World News API to get a news article. Make sure you have registered an account, logged in to the console and retrieved the API key.
We will use the search-news endpoint to find and filter an article that we are interested in. In this example, we will find an article related to Tesla, the electric car company.
Call the World News API using the following cURL command. Make sure to replace NEWS_API_KEY with your API key.
curl -L 'https://api.worldnewsapi.com/search-news?language=en&source-country=us&text=tesla&news-sources=https%3A%2F%2Farstechnica.com&sort=publish-time&sort-direction=DESC' \
-H 'x-api-key: NEWS_API_KEY'The request above sets the following parameters:
languagetoenfor English language newssource-countrytousfor news from the United Statestexttoteslato search for articles related to Teslanews-sourcestohttps://arstechnica.comto search for articles from Ars Technicasorttopublish-timeto sort the results by publish timesort-directiontoDESCto sort the results most recent first
You should receive a response similar to below, which includes an array of relevant news articles:
{
"offset": 0,
"number": 10,
"available": 7,
"news": [
{
"id": 233596680,
"title": "Breaking: Elon Musk drops claims that OpenAI abandoned mission",
"text": "While Musk has spent much of today loudly criticizing the Apple/OpenAI deal, he also sought to drop his lawsuit against OpenAI, a court filing today showed. In the filing, Musk's lawyer Morgan Chu notified the Superior Court of California in San Francisco of Musk's request for dismissal of his entire complaint without prejudice. There are currently no further details as to why Musk decided to drop the suit. Musk's lawsuit accused OpenAI CEO Sam Altman and President Greg Brockman of drifting away from its original nonprofit mission by forming a for-profit subsidiary later. It also claimed that OpenAI breached its founding agreement to make notable AI advancements freely available to the public. Musk co-founded OpenAI in 2015 and previously provided funding but is no longer with the company after leaving in 2018. \"To this day, OpenAI Inc.'s website continues to profess that its charter is to ensure that AGI ‘benefits all of humanity,'\" the lawsuit read. \"In reality, however, OpenAI Inc. has been transformed into a closed-source de facto subsidiary of the largest technology company in the world: Microsoft.\" Musk had previously requested a 10-day jury trial that sought an injunction and maximum punitive damages, but OpenAI had denied that he was entitled to any relief. \"The complaint fails to state a claim for breach of contract, as Musk cannot allege the existence of any 'Founding Agreement,'\" OpenAI had argued, calling Musk's complaint a \"fiction.\" \"He attaches no contract by that name to the complaint, and the documents he does attach contradict his allegations as to the alleged terms of the agreement,\" OpenAI argued. Musk's request for dismissal comes the day before the hearing where discovery was supposed to start in the lawsuit. OpenAI had objected to some of Musk's discovery requests for confidential information—with both sides arguing over terms of a protective order. Now, OpenAI has seemingly escaped from sharing any internal documents that it perhaps doesn't want Musk or the public to have access to. Musk has so far not posted about his request to dismiss the lawsuit on X, seemingly more focused on attacking the OpenAI/Apple deal. Previously, OpenAI had released a blog clarifying the company's relationship with Musk in light of the lawsuit. In it, OpenAI shared emails between its execs and Musk, while confirming that Musk had suggested that OpenAI should “attach to Tesla as its cash cow.\" On X, Musk threatened to ban Apple devices at Tesla and his other companies over the deal, which some users interpreted as Musk being jealous of the OpenAI deal.",
"summary": "Musk previously hoped a jury would award maximum punitive damages.",
"url": "https://arstechnica.com/tech-policy/2024/06/breaking-elon-musk-drops-claims-that-openai-abandoned-mission/",
"image": "https://cdn.arstechnica.net/wp-content/uploads/2024/06/GettyImages-2148557371-760x380.jpg",
"video": null,
"publish_date": "2024-06-11 21:51:49",
"author": "Ashley Belanger",
"authors": [
"Ashley Belanger"
],
"language": "en",
"source_country": "US",
"sentiment": -0.2
}
...
]
}Each item in the array includes a title, text and image. For the purpose of this guide, we will use the first item in the array, titled “Breaking: Elon Musk drops claims that OpenAI abandoned mission”.
Summarize the news article with the OpenAI Responses API
The article text returned in the response is quite long and not written to be read aloud. We will use OpenAI to summarize the article into a short script that a newsreader voice can read in about 25 seconds.
Make sure you have registered an account with OpenAI and created an API key. Below is an example cURL request. The input holds the instructions followed by the text of the article from the World News API. Replace OPENAI_API_KEY with your OpenAI API key, and ARTICLE_TEXT with the article text.
curl https://api.openai.com/v1/responses \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer OPENAI_API_KEY' \
-d '{
"model": "gpt-6.1-sol",
"input": "Write a 25 second summary of the following news article for a news anchor to read aloud. Use plain sentences only: no headings, lists or stage directions.\n\nARTICLE_TEXT"
}'Asking for plain sentences matters: anything in the script is read aloud, including headings and notes. The response takes a few seconds and looks something like this (shortened):
{
"id": "resp_00fe85ab39a1778c006af1e69ee77487d2b6ff592d62b2d517",
"object": "response",
"status": "completed",
"model": "gpt-6.1-sol",
"output": [
{
"type": "message",
"role": "assistant",
"content": [
{
"type": "output_text",
"text": "Elon Musk has asked a California court to dismiss his lawsuit against OpenAI, without explaining why. The suit accused the company of abandoning its nonprofit mission and breaking a founding agreement, claims OpenAI denied. The request comes a day before a scheduled hearing. Musk sought dismissal without prejudice, meaning he could file again, while separately criticizing Apple’s new deal with OpenAI."
}
]
}
],
"usage": {
"input_tokens": 555,
"output_tokens": 78,
"total_tokens": 633
}
}The text field inside output contains the summary. That is the script for the voice over.
Build the news video edit
Now we describe the video in JSON. The summary becomes the voice over, the headline sits in a lower third, and a category tag goes in the top corner. Create a file named news.json and paste the following JSON into it:
{
"timeline": {
"background": "#000000",
"tracks": [
{
"clips": [
{
"asset": {
"type": "rich-caption",
"src": "alias://anchor",
"font": {
"family": "Roboto",
"size": 34,
"weight": 700,
"color": "#ffffff"
},
"stroke": {
"width": 3,
"color": "#000000"
},
"active": {
"font": {
"color": "#fcba03"
}
},
"animation": {
"style": "highlight"
}
},
"start": 0,
"length": "end",
"width": 1100,
"height": 130,
"position": "bottom",
"offset": {
"y": 0.19
},
"fit": "none"
}
]
},
{
"clips": [
{
"asset": {
"type": "rich-text",
"text": "Breaking: Elon Musk drops claims that OpenAI abandoned mission",
"font": {
"family": "Roboto",
"size": 30,
"weight": 700,
"color": "#ffffff"
},
"align": {
"horizontal": "left",
"vertical": "middle"
},
"padding": 24,
"background": {
"color": "#101820",
"opacity": 0.9
}
},
"start": 0,
"length": "end",
"width": 1280,
"height": 100,
"position": "bottom"
}
]
},
{
"clips": [
{
"asset": {
"type": "rich-text",
"text": "Technology",
"font": {
"family": "Roboto",
"size": 22,
"weight": 700,
"color": "#ffffff"
},
"style": {
"letterSpacing": 2,
"textTransform": "uppercase"
},
"padding": 14,
"background": {
"color": "#c62828",
"borderRadius": 6,
"wrap": true
}
},
"start": 0,
"length": "end",
"width": 400,
"height": 60,
"position": "topLeft",
"offset": {
"x": 0.03,
"y": -0.04
}
}
]
},
{
"clips": [
{
"alias": "anchor",
"asset": {
"type": "audio",
"prompt": "Elon Musk has asked a California court to dismiss his lawsuit against OpenAI, without explaining why. The suit accused the company of abandoning its nonprofit mission and breaking a founding agreement, claims OpenAI denied. The request comes a day before a scheduled hearing. Musk sought dismissal without prejudice, meaning he could file again, while separately criticizing Apple’s new deal with OpenAI.",
"model": "polly-neural",
"options": {
"voice": "Matthew",
"newscaster": true
}
},
"start": 0,
"length": "auto"
}
]
},
{
"clips": [
{
"asset": {
"type": "image",
"prompt": "Editorial news photo for the headline \"Breaking: Elon Musk drops claims that OpenAI abandoned mission\". No people, no text, no logos.",
"model": "nano-banana-2",
"options": {
"aspectRatio": "16:9"
}
},
"start": 0,
"length": "alias://anchor",
"effect": "zoomInSlow"
}
]
}
]
},
"output": {
"format": "mp4",
"resolution": "hd"
}
}The first track in the array is the top layer, so the tracks read from the front of the picture to the back:
- Captions: a
rich-captionclip with"src": "alias://anchor". It transcribes the voice over and shows each word as it is spoken, just above the headline. - Headline: a
rich-textclip with the articletitleon a dark bar across the bottom of the frame. - Category: a small
rich-texttag with a red background."wrap": truemakes the background fit the word. - Voice over: an
audioclip with the summary inprompt. It uses Amazon Polly’sMatthewvoice with"newscaster": truefor a newsreader delivery."length": "auto"makes the clip as long as the speech. - Background image: an
imageclip generated from the headline with thenano-banana-2model, with a slow zoom. Its length comes from the voice over through"alias://anchor".
The World News API also returns the article’s image URL. If you have the rights to use that photo, set it as src on the image asset instead of prompt, model and options.
Render the news video
Send the JSON to the Edit API. Replace SHOTSTACK_API_KEY with your Shotstack API key:
curl -X POST \
-H 'Content-Type: application/json' \
-H 'x-api-key: SHOTSTACK_API_KEY' \
-d @news.json \
https://api.shotstack.io/edit/v1/renderYou should receive a response with the render id. Replace RENDER_ID in the cURL command below with that id to check the status of the video render:
curl https://api.shotstack.io/edit/v1/render/RENDER_ID \
-H 'x-api-key: SHOTSTACK_API_KEY'The status is generating while the voice and the image are created, then rendering. Once the status is done, you can view or download the video from the url value:
{
"success": true,
"message": "OK",
"response": {
"id": "01afa452-ca6d-4f47-8e30-9f5be76b37cb",
"owner": "5ca6hu7s9k",
"plan": "payg",
"status": "done",
"error": "",
"duration": 23.02,
"billable": 23.02,
"renderTime": 8040.62,
"credits": 1.28,
"url": "https://shotstack-api-v1-output.s3-ap-southeast-2.amazonaws.com/5ca6hu7s9k/01afa452-ca6d-4f47-8e30-9f5be76b37cb.mp4",
"poster": null,
"thumbnail": null,
"created": "2026-11-09T15:31:34.787Z",
"updated": "2026-11-09T15:33:14.378Z"
}
}Our final automated AI news video looks like this:
Create news videos from your own code
To produce a video for every story, build the edit in code. Here is the Shotstack step in Node.js 18 or later, with no dependencies. buildNewsEdit takes a headline, a summary and a category, and render sends the edit and waits for the video:
// news-video.mjs: turns a news story into a narrated video with captions in one render.
// Run with: SHOTSTACK_API_KEY=your-key node news-video.mjs (Node 18 or later)
const API = 'https://api.shotstack.io/edit/v1';
const headers = {
'Content-Type': 'application/json',
'x-api-key': process.env.SHOTSTACK_API_KEY,
};
function buildNewsEdit({ headline, summary, category }) {
return {
timeline: {
background: '#000000',
tracks: [
{
clips: [
{
asset: {
type: 'rich-caption',
src: 'alias://anchor',
font: { family: 'Roboto', size: 34, weight: 700, color: '#ffffff' },
stroke: { width: 3, color: '#000000' },
active: { font: { color: '#fcba03' } },
animation: { style: 'highlight' },
},
start: 0,
length: 'end',
width: 1100,
height: 130,
position: 'bottom',
offset: { y: 0.19 },
fit: 'none',
},
],
},
{
clips: [
{
asset: {
type: 'rich-text',
text: headline,
font: { family: 'Roboto', size: 30, weight: 700, color: '#ffffff' },
align: { horizontal: 'left', vertical: 'middle' },
padding: 24,
background: { color: '#101820', opacity: 0.9 },
},
start: 0,
length: 'end',
width: 1280,
height: 100,
position: 'bottom',
},
],
},
{
clips: [
{
asset: {
type: 'rich-text',
text: category,
font: { family: 'Roboto', size: 22, weight: 700, color: '#ffffff' },
style: { letterSpacing: 2, textTransform: 'uppercase' },
padding: 14,
background: { color: '#c62828', borderRadius: 6, wrap: true },
},
start: 0,
length: 'end',
width: 400,
height: 60,
position: 'topLeft',
offset: { x: 0.03, y: -0.04 },
},
],
},
{
clips: [
{
alias: 'anchor',
asset: {
type: 'audio',
prompt: summary,
model: 'polly-neural',
options: { voice: 'Matthew', newscaster: true },
},
start: 0,
length: 'auto',
},
],
},
{
clips: [
{
asset: {
type: 'image',
prompt: `Editorial news photo for the headline "${headline}". No people, no text, no logos.`,
model: 'nano-banana-2',
options: { aspectRatio: '16:9' },
},
start: 0,
length: 'alias://anchor',
effect: 'zoomInSlow',
},
],
},
],
},
output: { format: 'mp4', resolution: 'hd' },
};
}
async function render(edit) {
const res = await fetch(`${API}/render`, { method: 'POST', headers, body: JSON.stringify(edit) });
const body = await res.json();
if (!res.ok) throw new Error(`Render request failed: ${JSON.stringify(body)}`);
const id = body.response.id;
while (true) {
await new Promise((resolve) => setTimeout(resolve, 5000));
const status = await (await fetch(`${API}/render/${id}`, { headers })).json();
const { status: state, url, error } = status.response;
console.log(`Render ${id}: ${state}`);
if (state === 'done') return url;
if (state === 'failed') throw new Error(`Render failed: ${error}`);
}
}
const url = await render(
buildNewsEdit({
headline: 'Breaking: Elon Musk drops claims that OpenAI abandoned mission',
summary:
'Elon Musk has asked a California court to dismiss his lawsuit against OpenAI, without explaining why. ' +
'The suit accused the company of abandoning its nonprofit mission and breaking a founding agreement, ' +
'claims OpenAI denied. The request comes a day before a scheduled hearing. Musk sought dismissal ' +
'without prejudice, meaning he could file again, while separately criticizing Apple’s new deal with OpenAI.',
category: 'Technology',
})
);
console.log(`Video: ${url}`);Save it as news-video.mjs and run it with your API key in the environment:
SHOTSTACK_API_KEY=YOUR_API_KEY node news-video.mjsIn your application, call buildNewsEdit with the title from the World News API and the summary from OpenAI, once per story.
Turn every news story into a narrated video
This guide showed how to turn a news article into a narrated video with captions, using a news API, an LLM summary and one Shotstack render. We ran each step with cURL, and the Shotstack step is already a function you can call for every story.
From here, you could run it on a schedule for the day’s top stories, or render one video per topic for each of your readers.
Sign up for a free Shotstack account and render your first automated news video.
Frequently asked questions (FAQs)
How much does each news video cost?
The 23-second video in this guide used 1.28 credits: rendering, the newsreader voice and the generated background image together. The render status response shows this total in its credits field. To price a voice or an image before you render, send the asset to the POST /generate/quote endpoint, which returns the cost without charging anything.
Can I change the newsreader’s voice or language?
Yes. Change voice in options, or change model to an ElevenLabs or MiniMax speech model and choose one of its voices. Newscaster delivery is an Amazon Polly option that works with the Matthew, Joanna, Lupe and Amy voices. The AI speech generation guide lists every model, voice and language option.
How long does a news video take to render?
The video in this guide took about 100 seconds from request to finished file, including generating the voice and the image. Assets generated before are reused at no charge, so rendering the same story again doesn’t pay for the voice or the image a second time.
If you render many stories at once, a webhook tells your application when each video is ready, instead of polling.