Turn Three Articles into a Narrated News Bulletin

AI writes the scripts, the opening text, the three headlines of the rundown and the post copy. AI also records the voiceovers and makes the music. The layouts, the titles and the cards come from templates.

The bulletin plays the opening first. Then it plays the stories in the order of your links, with a badge that shows Story 1 of 3, Story 2 of 3 and Story 3 of 3.

Yes. The workflow makes a separate clip for each story, with its own title, captions and music. The clip is the same story as in the bulletin.

The prompts ask for only facts from the article, and for no opinion. AI can still make a mistake, so check each number and name before you publish.

The example run on this page used 124 API credits and 18.8 AI credits. The cost of a run can change with your input, for example the length of a video or the number of rows in a spreadsheet. There is no charge for the workflow itself.

Yes. You do not have to open the app to run it. You can start it from your own code with the API, from a Zap in Zapier, or from an AI assistant that is connected to the Bannerbear MCP server. Each run takes your data and gives you the finished files.

Yes. The install gives you your own copy of the workflow and of its templates. You can change the templates, such as the colors, fonts and logo, and you can change the steps of the workflow.

A run is not instant, so the API answers at once, and the run continues in the background. Ask for the run to see its status. The status is queued, running, completed or failed. When the run is completed, it lists its outputs.

A sequence of steps that runs as one thing. Each step is an image render, an animation render, or a tool — and a step can use the output of an earlier step as its input.

Define its inputs, then supply them when you run it. Steps reference those inputs the same way they reference each other’s outputs.

POST to the workflow’s run endpoint with its inputs. The run is queued and reports progress as each step completes.

That is the point of inputs. Define the parts that change — a headline, an image URL, a name — and supply them per run.

Yes. Each run records every step, its inputs, its outputs and how long it took, so a failure points at the step that caused it rather than the whole workflow.

The run stops there and the failure is recorded against that step, with the outputs of the steps that already succeeded still available.

As long as its steps need — video work is not instant, and a workflow with several video steps takes minutes rather than seconds. Use the run’s progress or a webhook rather than holding a request open.

The workflow and its bundled templates are copied into your workspace, and the steps are rewritten to point at your copies rather than the originals.

Three on a trial. They are counted per workspace.

A workflow decides what gets made. An integration decides what triggers it and where the result goes. They are often used together.