Can ChatGPT Make Videos? Yes, With a Connector: How to Generate and Post Reels From ChatGPT

Short answer: ChatGPT cannot render a short-form video by itself. It can write a script, a shot list and a caption for one, and it can describe a video in loving detail, but the model produces text. What changed in 2025 is that ChatGPT can now call outside tools through connectors, and one of those tools can be a video renderer. Connect Shortzly, and "make me a 30-second reel about compound interest and post it to TikTok" becomes a request ChatGPT can actually complete.

This guide covers what ChatGPT does natively, what a connector adds, the exact steps to connect Shortzly, prompts that produce good reels, and what it costs.

What ChatGPT can and cannot do with video today

Out of the box, ChatGPT is very good at the writing half of a video: hooks, scripts, shot lists, on-screen text, descriptions and hashtags. Ask it for "a 6-segment script for a faceless reel about the Roman aqueducts, punchy, under 80 words" and you will get something usable on the first try.

It does not, on its own, produce an MP4 with a voiceover, stock footage, motion and burned-in captions, and it does not post anything to TikTok, YouTube or Instagram. OpenAI's separate video model, Sora, generates clips from a prompt, which is a different job from a scripted, narrated, captioned reel built from your own product screenshots and published to your accounts.

So the honest answer to "can ChatGPT make videos" is: the writing yes, the rendering and publishing only with a tool. Connectors are how you give it that tool.

What a connector is

ChatGPT connectors are built on the Model Context Protocol (MCP), an open standard for letting AI assistants call external tools. A remote MCP server publishes a list of tools with JSON schemas. When you add one as a connector, ChatGPT reads the list and can call the tools during a conversation. The same standard works in Claude, Claude Code, Codex, Cursor and Windsurf, so a server built once serves all of them.

Shortzly runs a remote MCP server at https://shortzly.com/mcp with 14 tools: discover capabilities and quota, upload your own media, create a reel, track the render, fetch the clip, upgrade it to HD and publish it. The full list is on the MCP server page.

How to connect Shortzly to ChatGPT

1. Create a Shortzly account if you do not have one, and connect at least one social account under Social in the dashboard if you want ChatGPT to publish for you.
2. Open ChatGPT settings and go to Connectors.
3. Turn on Developer mode. This is the switch that allows custom MCP servers.
4. Click Create, name the connector Shortzly, and paste https://shortzly.com/mcp as the server URL. Leave authentication on OAuth.
5. Sign in when ChatGPT opens the Shortzly consent screen. The screen shows which host will receive the token, so you can confirm it is ChatGPT.
6. Enable the connector in a chat from the tools menu, then ask for a reel.

You can see and revoke the connection at any time under Settings, API tokens and MCP in Shortzly. Revoking there disconnects ChatGPT immediately.

Prompts that produce good reels

ChatGPT is writing the script, so the prompt matters more than it does in a template-based generator. Three patterns that worked well in our testing:

1. The explainer

Make a 30-second vertical reel explaining why the sky is blue.
Six segments, one sentence each, open with a question.
Nova voice, CapCut captions, stock footage. Preview quality first.

2. The product ad with your own screenshots

I will upload 5 screenshots of my app. Write a 40-second ad,
one screenshot per segment, end on our logo. Use the Onyx voice.
Render a preview, show me the download link, and wait for my approval
before upgrading to HD or posting.

ChatGPT will call create_upload for each file, give you the upload URLs or handle the upload itself depending on the client, then reference the uploaded assets in create_reel.

3. The hook test

Write 5 different opening lines for a reel about intermittent fasting.
Keep the same 4-segment body. Render all 5 as previews and list them
with their download links so I can pick one to upgrade to HD.

Previews cost half a credit each, so five variations cost less than three HD renders.

What happens behind the scenes

When you send the first prompt, ChatGPT usually calls get_capabilities to learn the valid voices, caption styles and limits, then get_account to check your remaining quota. It writes the script and calls create_reel with an array of segments. Each segment has the spoken line and one media source: a stock search phrase, a public URL or an uploaded asset reference.

Shortzly queues the render. ChatGPT polls get_job until the status is completed, then calls get_clip for a download and thumbnail URL. If you approve, upgrade_clip_hd re-renders at 1080x1920 and publish_clip posts it to the accounts you name, now or at a scheduled time. get_social_post returns the live link.

Shortzly never runs its own script AI on this path. That is deliberate: you are already paying for ChatGPT's writing, so Shortzly only charges for the render. Text-to-speech, Whisper caption alignment and B-roll detection are optional and switched on per call.

Tips for better results

• Ask for a preview first. Preview renders are fast and cost half a credit. Upgrade to HD only when the script and visuals are right.
• Give it a length and a segment count. "30 seconds, six segments" produces tighter pacing than "short".
• Name the voice and caption style. Nova and Shimmer read warm, Onyx and Echo read authoritative. CapCut and Hormozi styles suit fast talking; Minimal and Podcast suit calm explainers.
• Use your own media for anything branded. Stock footage is fine for explainers. For a product, upload screenshots and let ChatGPT place one per segment.
• Bring your own voiceover if you have one. Upload an audio file and choose timed captions, and the render uses none of Shortzly's AI, which keeps the credit cost at the render itself.
• Tell it to wait for approval before publishing. ChatGPT will otherwise do exactly what you asked, including posting.

What it costs

Nothing extra. The Shortzly MCP server is included on every plan and shares your monthly faceless reel quota: a preview render is half a credit, an HD render one credit. Starter ($19/month) includes 10 faceless reels a month, Pro ($49/month) includes 100. Read-only tools such as checking a job or listing clips are free. Your ChatGPT usage is billed by OpenAI as usual.

ChatGPT versus Claude versus Claude Code for this

All of them connect to the same server and see the same 14 tools. In practice the differences are about workflow, not capability. ChatGPT and Claude are the most comfortable for one-off reels in a conversation. Claude Code and Codex are better when you want to script the process, for example rendering a reel for every row in a spreadsheet, because they run in your terminal and can loop. Cursor and Windsurf are useful when the reel is part of a larger project, such as generating a launch video from the same repository that holds the app.

If you want to see the terminal side, read how to add an MCP server to Claude Code, Codex, Cursor and Windsurf.

Frequently asked questions

Is this the same as ChatGPT plugins?

No. Plugins were retired. Connectors are the current mechanism and are built on MCP, an open standard that works across assistants.

Do I need a developer account?

No. Developer mode is a setting inside ChatGPT that allows custom connectors. You do not need an OpenAI API key or any code.

Can ChatGPT post the video for me?

Yes, to any TikTok, YouTube, Instagram, Facebook or LinkedIn account you have connected in Shortzly. Ask it to wait for your approval first if you want a review step.

Can I use my own voice?

Yes. Upload a voiceover file through create_upload with the voiceover purpose and choose timed captions. Shortzly lays the visuals and captions over your audio.

What if the render fails?

get_job returns the error and a retry hint, and create_reel accepts an idempotency key, so a retried request never renders twice. Failed renders are not charged.