Sunra MCP puts Seedance 2.5 inside Claude and ChatGPT from $0.039 a second, with no markup
Sunra has launched an MCP that brings leading video, image and audio models into the assistants you already use, priced at each maker's own API rate. Here is what it could change for AI filmmakers, and what to check first.
If you have ever kept a spreadsheet of video model prices open in one tab while generating in another, Sunra has just aimed a product at you. Sunra MCP puts image, video and audio models inside the AI assistants you already work in, including Claude, ChatGPT, Cursor and Codex. Seedance 2.5 starts at $0.039 a second.
The pitch is simple. Each model is priced at its maker's own API rate, with no markup, per creator @kevindatas. Creator @Just_sharon7 says Sunra is public about every price, even where it is not the cheapest, and quotes GPT Image 2.5 from $0.004 an image. One balance reportedly covers both the assistant connection and the Sunra app. For anyone who bounces between a chat window and a generation dashboard all day, that is a real change in how the work flows.
What is actually being claimed
Three independent creators describe the same launch, which is why this is written with attribution. Per @dr_cintas, every leading image, video and audio model is now reachable from inside Claude and ChatGPT at each maker's own API price. The shared figures match across all three posts: Seedance 2.5 from $0.039 per second, GPT Image 2.5 from $0.004 per image.
What we have not seen is a post from Sunra itself in this batch. Treat the pricing as reported, and check Sunra's own page before you budget a project around it.
What it changes in a real workflow
The useful idea is not the price. It is that generation becomes a step inside a conversation. You can ask an assistant to draft a shot list, then have it call a video model for each shot, then ask it to revise the prompt on shot four. No copying prompts between tabs, no re-uploading reference frames by hand.
Compare that with the current alternative. Most of us either use a platform's own web app, which is convenient but locks us to its model list and its credit pricing, or call a model API directly, which is cheap but needs code. An MCP sits between the two. It claims API pricing with chat-level convenience.
At $0.039 a second, a 10-second Seedance 2.5 clip would come to about $0.39 at the quoted starting rate. That is simple arithmetic on a starting price, not a quote for any particular resolution or mode. Higher settings may cost more.
What was not said
There are gaps worth naming:
- The "from" in the price. Starting rates rarely apply to every output option. Resolution, duration and audio could all move the number.
- Which models are included. "Every leading model" is a broad phrase. No list appears in these posts.
- Limits and queueing. Nothing is said about rate limits, concurrency or how long a clip takes to come back inside a chat.
- Rights and terms. We do not know how commercial usage terms pass through from each model maker.
- Availability date. The posts say it has launched. No rollout schedule or regional notes were given.
The no-markup claim is also a business model question. Sunra has to earn money somewhere, and none of the posts say how.
What to test first
Start small. Connect the MCP to the assistant you already use and run one shot you have already made elsewhere, with the same prompt and the same reference image. Compare cost per second, time to result and how closely the output matches.
Then test the part that is new. Ask the assistant to generate a three-shot sequence from one brief and see whether it keeps character and tone consistent without you stepping in. If it does, that is the saving that matters, because your time costs more than the credits.
Finally, check the invoice. Run ten clips, then compare the balance you spent against the published per-second rate. If the numbers match, the no-markup claim holds up. If they do not, you will want to know before a client project depends on it.