Create a Branded Carousel with ChatGPT and Prolifik

Set the delivery and the audience

The intended reader is a small-team office manager. They need a practical reminder they can use before a hybrid meeting. The message is not a product tour, so the images can illustrate the setting without pretending to show a specific app or meeting room.

Tell ChatGPT: “Prepare one five-slide carousel about welcoming a remote guest into a hybrid meeting. Each slide should give one concrete action. Use our saved brand, and return rendered media for review. We will decide the posting destination after seeing it.”

For a more detailed explanation of the formats, use the image-versus-video carousel guide. A video version may need shorter wording than a set of swipeable images. Do not assume the destination will interpret an MP4 as an image carousel.

Give every slide a distinct role

Here is an illustrative five-slide sequence:

1. “Your remote guest should not have to interrupt to join the room.”
2. “Place the microphone where the quietest speaker can be heard.”
3. “Open the shared document before the meeting starts.”
4. “Name who will watch the chat while someone presents.”
5. “Before you begin, ask the remote guest what they can hear and see.”

The opening identifies the problem. Each middle slide provides a different practical action. The ending brings the advice into the actual meeting. Replacing every middle slide with a broad phrase such as “improve communication” would remove the value of the carousel.

Ask ChatGPT to preserve the five lines exactly after you settle them. A separate post caption can add context, but it should not silently become a sixth slide. The carousel caption guide helps keep those two writing tasks separate.

Pass slide copy in the right shape

Prolifik's create_content_job tool accepts format: carousel. For one carousel, the count is one, even when it contains five slides. The customTexts field holds one object whose slides array contains the five strings.

This is an illustrative fragment of that request, not a complete call:

{
  "format": "carousel",
  "count": 1,
  "deliveryMode": "download",
  "customTexts": [{
    "slides": [
      "Your remote guest should not have to interrupt to join the room.",
      "Place the microphone where the quietest speaker can be heard.",
      "Open the shared document before the meeting starts.",
      "Name who will watch the chat while someone presents.",
      "Before you begin, ask the remote guest what they can hear and see."
    ]
  }]
}

The full request also needs the actual selected brand identifier and an idempotency key. ChatGPT should obtain the brand from the connected account, not invent an identifier. One object per carousel also matters in a batch: five objects would request five pieces, not five slides in one piece.

Choose visuals that help the lesson

The tool supports carousel library filters such as a theme. A general workplace theme may suit this example, but it cannot guarantee a particular meeting-room photograph. If a specific real room, screen or product matters, verify that the chosen workflow accepts and uses that supplied media.

Do not let an assistant label stock imagery as your office or a real customer session. The copy already carries the practical lesson. Suitable general imagery can establish context without pretending to document an event.

The Image Carousel format page explains the visual form. Keep logos and text clear of busy image areas, and inspect the actual result rather than judging only the planned palette or background description.

Retrieve all of the returned media

After submission, keep the content job identifier and retrieve the job with get_content_job. Rendering continues asynchronously. The response provides pieces with their script and media fields, including slide URLs or a video URL when available.

Ask ChatGPT to report what actually exists: “Show the returned piece ID, the number of rendered slide images, their order and any video output. Do not describe missing images as ready.” This catches a common handoff problem where a video preview is mistaken for five downloadable images.

If one slide needs a wording change, update_content_piece supports a replacement slides array and starts another render. Retrieve the piece again after it becomes ready. Keep the final ordered set together; mixing old and new slide versions makes the carousel inconsistent.

Inspect the sequence before selecting a destination

Open every slide at phone width. Read the longest line aloud, check contrast and confirm the first and last slides match their roles. On the meeting-room example, slide four should still say who watches the chat, rather than collapsing into a generic reminder to pay attention.

Then inspect any video version without pausing. If viewers cannot finish the text before it advances, shorten the copy or choose image delivery where supported. Native image carousel publishing depends on both the returned media and the destination path; it is not universal across social platforms.

Use the Prolifik MCP guide to connect the service and inspect its current tool options. OpenAI's connection guidance covers ChatGPT availability and permissions. Once the carousel passes review, make a specific publishing request that names the media, account and intended delivery instead of asking the assistant to “post the carousel everywhere.”