# Luma's Connector Puts Its Video and Image Workflows Inside Claude and ChatGPT: What It Does and What It Costs

- Published: 9 November 2026
- Updated: 9 November 2026
- Author: RenderBob team
- Category: Product
- Canonical: https://renderbob.ai/blog/luma-connector-claude-chatgpt-creative-workflows

Luma's connector, announced 3 November, runs its image and video workflows inside Claude or ChatGPT and bills your Luma credits. We have not tested it.

Luma announced on 3 November a connector that lets teams run its image and video workflows from inside Claude or ChatGPT. The details come from Luma's own pages, which are promotional, and the connector uses your Luma credits.

## What Luma says it does

Luma's announcement and a connector guide dated 19 October describe the following.

- Single generations. One request returns one result. Luma names Uni-1 for images and Ray 3.2 for video, plus other models a team can run. Examples include stills, 5- or 10-second clips and image edits.
- Mini apps. Multi-step workflows that plan, generate, check their own output and return assets. In Claude they include Ad Variants (five aspect ratios, five languages, with a preview before final renders), Background Swap, Camera Angles, Light Transfer (up to 4K) and Face Swap. Product Shots, Product 360, Virtual Try-On and Layers are in the Luma app only.
- Asset library. Claude can browse the team library, use files as inputs and save results. It deletes files only when asked.
- Usage. Claude can report live credit usage and session costs.

## Setup, requirements and cost

The guide says you add the connector in Claude's settings, sign in to Luma and pick a team with credits. Generations bill to that team at app rates, with no separate subscription. It says this often fails for personal teams without credits. Teams on Claude Team or Enterprise can add it for a whole workspace. A Luma plugin offers similar setup in ChatGPT. No per-generation credit costs are given.

## Limits worth knowing

- Video can take several minutes.
- Results do not appear on Luma boards automatically. They must be saved to the asset library and imported.
- Anything saved there is shared with the team.
- Claude may ask permission before each step unless you allow it.
- Re-running a stuck video may charge twice, so check whether it is ready first.
- Luma's page gives no output length, resolution or regional limits beyond those above.

## What is not shown

The announcement is promotional and gives no benchmarks for output quality or speed. It cites a statistic about how often advertisers refresh creative, which says nothing about Luma's results. We have not tested the connector.

## A sensible trial

- Use a test team with a small credit balance.
- Run one real brief through a single generation and one mini app, and record time and credits.
- Check the output rules for licence, likeness and library sharing.
- Compare cost per usable asset with your current tool, as in our 3 November guide.
- Keep a record of each prompt and result outside the chat.

## What to watch

Luma's credit prices per workflow, independent tests of the mini apps, and whether more generation services appear as connectors inside chat assistants.

## RenderBob 3 and this story

RenderBob 3 starts from the editor: generated clips sit on their own timeline lane, the model is a per-clip choice, and references and prompt logs stay with the studio.