--- title: "How we make our videos with an AI agent | Kylon" description: "How the Kylon team makes its videos: reviewers annotate the cut on the frame, decisions go into the agent's video-style skill, built-in GPT Image, Seedance and ElevenLabs remake it, and the site change goes live on approval." url: "https://kylon.io/use-cases/how-we-run-content" language: "en" --- Connection series · From first video cut to live on the site # How we make our videos. Reviewers annotate the first cut right on the frame and send their notes to the thread. When they disagree, Grace decides and Vale saves the call to its video-style skill. Vale remakes the cut with GPT Image, Seedance and ElevenLabs, built into Kylon, and the homepage change waits as a draft until Mira clicks Publish. [Start free](https://app.kylon.io/sign-up) [Book a demo](https://kylon.io/demo?from=workflow-hero&page=%2Fen%2Fuse-cases%2Fhow-we-run-content) ## Who runs it - **Grace**, video owner. Decides when reviewers disagree. - **Lucas, Eli, Kenji**, reviewers. Annotate the cut on the frame. - **Mira**, approver. Publishes the website change. - **Vale**, agent. Saves each call to its video-style skill, remakes the cut with the built-in models, files it and drafts the page. ![Google Drive](https://kylon.io/v2/integrations/google-drive.png) ![GPT Image](https://kylon.io/v2/integrations/openai.png) ![Seedance](https://kylon.io/v2/integrations/seedance.png) ![ElevenLabs](https://kylon.io/v2/integrations/elevenlabs.png)1 connection · 3 built in This is how the Kylon team runs it. The people in the video are illustrative. [All six films on YouTube](https://www.youtube.com/playlist?list=PLTv_r3jinR0U) Connections ## What each tool does here. Each person signs in to their own account. The agent works with what that account can already see. Step 05 **Google Drive** The final film goes into the team's Drive folder, named and dated. Step 04 · Built in **[GPT Image](https://kylon.io/integrations/gpt-image2)** Makes the new background the reviewers asked for. Built into Kylon, nothing to connect. Step 04 · Built in **[Seedance](https://kylon.io/integrations/seedance)** Regenerates the shot that needed changing. Built into Kylon, nothing to connect. Step 04 · Built in **[ElevenLabs](https://kylon.io/integrations/elevenlabs)** Writes a new score for the shorter cut. Built into Kylon, nothing to connect. The workflow ## 6 steps, one Room. The same steps as the film. Each one shows what the agent does in the Room and, where it matters, who signs it off. 1. _01_First cut ### First cut posted In Kylon **Vale**In the Room, playing in the video preview. 2. _02_Annotations ### Annotated on the frame In Kylon **Vale**Reviewers box what they mean and send notes to the thread. Who decides **Grace decides.** One call settles it. 3. _03_Skill rules ### Saved to the skill In Kylon **Vale**Vale writes the call into its video-style skill. 4. _04_Generated ### Remade with built-in models In Kylon **Vale**GPT Image for the background, Seedance for the shot, ElevenLabs for the score. 5. _05_Draft ### Final cut on Drive In Kylon **Vale**And the homepage change as a draft. Who decides **Mira checks the draft.** She clicks Publish. 6. _06_Live ### Live on the website In Kylon **Vale**The homepage opens with the new film. **Every call, kept**Two reviewers disagreed on the background and Grace decided. Vale saved it to its video-style skill, so the next film starts from it. Set it up ## Four steps to your own version. 1. 1**Connect your tools**In Connections, sign in to Google Drive. GPT Image, Seedance and ElevenLabs are already built in. 2. 2**Open a content Room**Add your agent, your reviewers and whoever approves the website. 3. 3**Paste the set-up prompt**The agent asks for your rules one at a time. 4. 4**Post the first cut**Reviewers annotate it on the frame. Every decision goes into the video-style skill. Set-up prompt Set up our content workflow in this Room. 1. When I post a cut, collect the annotations people send to the thread. If two notes disagree, ask the owner to decide. 2. Save every decision in a video-style skill, so the next film starts from it. 3. Make the changes with the built-in models: GPT Image for backgrounds and stills, Seedance for shots, ElevenLabs for music. 4. Put each final file in our Google Drive folder, named and dated. 5. Prepare any website change as a draft. It goes live only when the approver publishes it. Before you start, ask me for our rules one at a time. [Use this in Kylon](https://app.kylon.io/launch/workflows:how-we-run-content) ## Run it in your own workspace. Connect the tools you already use, add your team, and paste the prompt. Your agent asks before anything goes out. [Start free](https://app.kylon.io/sign-up) [Book a demo](https://kylon.io/demo?from=workflow-closing&page=%2Fen%2Fuse-cases%2Fhow-we-run-content) ## More workflows [**From customer inquiry to closed deal**](https://kylon.io/integrations/workflows/sales-inquiry-to-deal) [**From project files to tasks with owners**](https://kylon.io/integrations/workflows/project-delivery) [**From client welcome to contract renewal**](https://kylon.io/integrations/workflows/client-care) [**From daily ad report to budget changes**](https://kylon.io/integrations/workflows/how-we-run-ads) [**From finding creators to paying them**](https://kylon.io/integrations/workflows/how-we-run-creators) [Browse all integrations](https://kylon.io/integrations)