Event banners made with AI art and a Bannerbear template

Event banners made with AI art and a Bannerbear template

An n8n workflow with a form for an event's title, place, date and a prompt for its picture. OpenAI draws the picture, Cloudinary hosts it, and Bannerbear places it and the event text on a template. The finished banner is posted to a Discord channel.

What is inside

• workflow.json: the workflow as n8n exports it, fifteen nodes: seven that do the work and eight sticky notes. It holds no credential.
• README.md: these notes.

What we changed from the original copy

• The credential links are gone. Select your own for OpenAI, Bannerbear and Discord. The Cloudinary step needs none.
• The picture step was the OpenAI node at version 1.3, which draws with DALL·E 3 unless told otherwise, and the file names no other model. OpenAI's deprecations page, read on 12 November 2026, lists DALL·E 3 as removed on 12 May 2026. The node is now version 2.3 and names gpt-image-2.6-flare. The deprecations page lists gpt-image-1 for shutdown on 23 October 2026 and gpt-image-2 on 10 February 2027, and names the 2.5 and 2.6 sunburst and flare models as replacements.
• The options are 1024x1024 and quality medium. The old quality, standard, is a DALL·E 3 value. The node still returns the picture as the binary field data, which the Cloudinary step reads.
• The Cloudinary address holds your cloud name as <YOUR_CLOUDINARY_CLOUD_NAME>, and the upload preset is <YOUR_CLOUDINARY_UPLOAD_PRESET>. The step linked a Query Auth credential. Cloudinary's documentation says an unsigned upload needs only file and upload_preset, with no signature or api_key, so the node's authentication is set to none. The preset must be unsigned.
• The form offered two Bannerbear templates by the author's own names, each tied to the author's template id. The choices are Template one and Template two, tied to <YOUR_BANNERBEAR_TEMPLATE_ID_ONE> and <YOUR_BANNERBEAR_TEMPLATE_ID_TWO>.
• Download Banner read image_url_jpg. Bannerbear's version 2 reference lists image_url only, so the node reads image_url_jpg when it exists and image_url otherwise.
• The Discord server and channel are <YOUR_DISCORD_SERVER_ID> and <YOUR_DISCORD_CHANNEL_ID>. The file option names the field data outright.
• The form's address was a random identifier. It is event-banner now, and the form is titled Event banner maker.
• The notes are rewritten. One showed a picture on the author's own Cloudinary account, and it is removed. Another said the banner goes to X, but the workflow posts to Discord, and the notes now say Discord.
• Nothing else.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An OpenAI API key.
• A Cloudinary account with an unsigned upload preset.
• A Bannerbear account, a template whose layers are named placeholder_image, placeholder_text, placeholder_location and placeholder_date, and a project API key for Bannerbear's version 2 API. Its documentation says a version 5 key does not work with version 2 endpoints.
• A Discord bot that is on your server and may post in the channel.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Add an OpenAI credential and select it in Generate AI Banner Image.
3. In Upload to Cloudinary, replace <YOUR_CLOUDINARY_CLOUD_NAME> in the address and <YOUR_CLOUDINARY_UPLOAD_PRESET> in the query.
4. In Set Parameters, replace the two template ids. Rename the two choices in the form if you like, and keep the two sets of names the same.
5. Add a Bannerbear credential and select it in Send to Bannerbear Template. If your layers have other names, change them in its modifications.
6. Add a Discord bot credential and select it in Discord. Replace <YOUR_DISCORD_SERVER_ID> and <YOUR_DISCORD_CHANNEL_ID>.
7. Open the form's test address, fill it in and submit.

How a run goes

1. n8n Form Trigger shows a form with a template choice and the title, place, date and image prompt.
2. Set Parameters turns the template choice into a template id and copies the other fields.
3. Generate AI Banner Image draws the picture from the prompt.
4. Upload to Cloudinary posts the file with the upload preset. The answer holds secure_url.
5. Send to Bannerbear Template fills four layers: the picture, taken from secure_url with f_auto,q_auto added, then the title, the place and the date. It waits for the image, trying up to ten times.
6. Download Banner fetches the finished banner.
7. Discord posts the message New Event Alert! with the title, place and date, and attaches the banner.

Change the look

• The layout: the Bannerbear template, and the layer names in Send to Bannerbear Template.
• The picture: the prompt on the form, and the size and quality in Generate AI Banner Image.
• The message: the text in Discord.
• Where it goes: swap Discord for another social network's node.
• The file host: replace Upload to Cloudinary with another host that gives back a web address.

Good to know

• n8n's Bannerbear node calls the version 2 API and polls every two seconds, up to the number of tries you set, which is ten. A banner that takes longer ends the run with an error.
• An unsigned preset's name can be found in a web page's code. Cloudinary's documentation advises allowing only the formats you need with allowed_formats and turning on disallow_public_id.
• Each run costs one OpenAI image, one Cloudinary upload and one Bannerbear image.
• This copy was not run against OpenAI, Cloudinary, Bannerbear or Discord.
• The prompt goes to OpenAI, the picture to Cloudinary and Bannerbear, and the banner to Discord.
• The JSON holds no credential.