Take a video from thumbnail to published with promo posts
A walkthrough of the pipeline I built with Claude Cowork: I drop in a video, Claude handles distribution and promotion, and I stay in the loop as editor. Task: The author built a six-phase pipeline using Claude Cowork to automate video-to-published workflows. Inputs include raw video files dropped into a project directory. Claude handles transcription via Whisper, drafts blog posts, generates thumbnails in Figma, and schedules social media posts. The human acts as an editor, reviewing proposals, providing feedback via voice notes, and approving final outputs before publication.; tools: Claude Cowork. Human role: The original post does not say how a person stayed involved.; limitations: The workflow requires local execution of Whisper for transcription and access to Figma for thumbnail generation. Bluesky scheduling relies on custom tasks due to lack of native support. The author notes that the process is not yet fully streamlined, requiring human intervention at every phase to maintain voice authenticity and correct errors..
Tools
Published on Agent Work Atlas:
Overview
A walkthrough of the pipeline I built with Claude Cowork: I drop in a video, Claude handles distribution and promotion, and I stay in the loop as editor.
Workflow
The author built a six-phase pipeline using Claude Cowork to automate video-to-published workflows. Inputs include raw video files dropped into a project directory. Claude handles transcription via Whisper, drafts blog posts, generates thumbnails in Figma, and schedules social media posts. The human acts as an editor, reviewing proposals, providing feedback via voice notes, and approving final outputs before publication.
Reported result
The author reports that the process currently takes more than the one-hour goal due to ongoing tweaks, but feels the foundation is strengthening. No specific metrics on time saved or content volume are provided. The workflow is described as still experimental, with the author noting they are 'quite a ways from proving the hypothesis' of daily publishing.
Human role
The original post does not say how a person stayed involved.
Limitations
The workflow requires local execution of Whisper for transcription and access to Figma for thumbnail generation. Bluesky scheduling relies on custom tasks due to lack of native support. The author notes that the process is not yet fully streamlined, requiring human intervention at every phase to maintain voice authenticity and correct errors.
Try it
Act as a content automation agent using Claude Cowork. I will provide a raw video file [replaceable input: video path] and a style guide [replaceable input: style guide URL/text]. Execute the following steps: 1. Transcribe the video using Whisper. 2. Propose titles and summaries for YouTube and a blog post. 3. Draft the blog post and create a thumbnail proposal in Figma. 4. Generate social media drafts for X, LinkedIn, and Bluesky. Stop after each step and wait for my explicit approval before proceeding. Do not publish anything without my confirmation. If a step fails, report the error and suggest a fix. Measure success by whether the final output matches my voice and style guide. If inputs are missing, ask for them before starting.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.