A weekly social clip list is a short table: for each recording from the week, the moments worth posting, each with a time range, a reason and a suggested angle for the post. Claude can produce it the same way every week if it can search the recordings; without that, it works only from transcripts or notes someone pastes in. The routine starts with getting the week's recordings in and ends with handing the picks to whoever cuts and posts, and the part in between is asking good questions.
Write the standing brief once
Save a short brief Claude can reuse each week: the formats you post (a guest's sharpest point, a customer describing their problem) and what rules a moment out (unreleased roadmap, anyone who hasn't signed a release). Add the columns you want back. The brief keeps week nine consistent with week two, and it's the kind of written habit that the reuse stage of a content workflow depends on.
Four ways to feed the week's recordings to Claude
Paste the transcripts and ask for quotable lines. It's the cheapest route and it works for spoken moments. A line that reads well can still play badly on camera, so someone watches every pick, and choosing the seconds before making the vertical clip is where that judgment sits.
Give Claude a filesystem MCP server on the recordings folder. It can see what arrived this week, including file names and any transcript files, which tells it what exists and nothing about what happens on screen.
Build a transcription and vector search pipeline over the back catalog. Every past episode becomes searchable by what was said. Someone has to maintain it, and moments that are only visible need more work.
Use a hosted video search connector that indexes picture and sound and lets Claude search over MCP, at the cost of an upload step and a usage allowance.
A different category skips the list altogether. Automatic clipping tools cut clips straight from signals like laughter or applause, and what those signals catch and miss decides whether that's enough for your channel.
Build this week's list from the recordings in a Claude chat
Vivu can be added to Claude as a custom connector, and the weekly routine starts with uploading the new recordings to the series project, where each one is indexed once in the cloud. Ask by description and put the subject into the question: "find a moment where a guest pushes back on another panelist about pricing" will probably hold up better than "find a disagreement", because the search can take a verb literally, and asking for a speaker who corrects another panelist can return someone clearing up a small mix-up about how the session was being run. Results come back as time ranges with a one-line reason each, and you open them on the results page.
Claude then writes the week's table with an angle for each clip, and the social editor cuts from the ranges the team kept.
When a weekly search routine doesn't fit
It won't suit a team that wants one search across every series and year, because every search is scoped to a single project, and it adds a weekly step, because recordings have to be uploaded to a Vivu project and indexed in the cloud before they can be found. Each week's searches also come out of the plan allowance (Vivu has a free tier and a paid tier, with the limits on its MCP page), and what you get are time ranges to open and trim, found by what happens in them, so check the exact wording of any quote against the transcript or the audio.
When you don't need any of this
One recording a week with a host who flags the good moments during the call doesn't need a search step. Ask the host for the flags and have Claude format them. If your team is content to post whatever an automatic clipping tool picks, a hand-built list adds a review step your channel may not need.
Which way your team should run it
Look at who chooses the clips today. If it's the person who hosted, they already know the moments, and Claude is a formatter. If it's a social manager who wasn't on the call, the list is how they learn what's in the recording, and building it from a search over the footage is worth the weekly upload. If nobody chooses and clips simply come out of a tool, decide whether that was ever a choice.
FAQ
Do I have to upload each week's recordings before Claude can search them?
Yes, with an upload-based video search connector. The new recordings go into the project first; after that, each file is indexed once and stays searchable in later conversations. Put the upload in the same weekly slot as the list so the two don't drift apart.
Why did a search for a moment in a recording come back empty?
Either the moment isn't in the recording or the description was stricter than the footage. A stage recording shot mostly wide may contain no close-up of a single speaker, so a request for one returns nothing. Loosen the wording once, or search for something you know is there, before deciding the moment doesn't exist.
Can I search the week's recordings for an exact quote?
Describe the moment instead of relying on exact wording. A description-based search is built around what happens in a recording, so for a post that quotes someone word for word, confirm the wording against the transcript or the audio before it's published.
Can Claude tell which clips will perform best?
Not reliably. Performance is decided by the audience after posting. Claude can apply the criteria in your brief, and if you paste in last month's post results it can look for patterns, but that is a guess about the audience and the list should label it as one.