Quickstart
Deploy Blockbuster Studio on Runpod, create your password, and make your first video.
This guide takes you from nothing to your first video, usually in about 15 minutes. You’ll need a Runpod account with some credit on it and a free Hugging Face account; What you need has the full list, including what to install for a previs or an AI agent.
Start the deploy
Go to blockbuster.studio and click Deploy on Runpod. There’s one studio with everything in it: films with consistent characters, props and locations, video with sound, character voices, and Perform. Its models (about 128 GB) download the first time it starts.
Deploy the pod
On Runpod’s deploy page, click the RTX PRO 4500 GPU card. It’s the card the studio is tested on. If it isn’t available, pick another card with 32 GB or more, like an RTX 5090.
Next, click Add volume and keep the suggested size (180 GB). This is where your models and films live. Don’t skip it: without a volume the models don’t fit, and your work is lost when the pod stops.
You don’t need to change anything under Set overrides. STUDIO_PASSWORD shows change-me, which the studio ignores: you’ll create your password in the studio. (If you want a smaller install, you can switch model downloads off here, under Set overrides; see Settings for the full list.)
Then click Deploy Pod.
Open your studio
When the pod is running, click Connect, then HTTP Service. The studio opens in a new tab.
Your studio’s address looks like https://<POD_ID>-3000.proxy.runpod.net. Bookmark it.
Create your password
On the first screen, choose a password of at least 8 characters and click Create password & enter.
Your studio is on the public internet, and this password is the only thing that keeps others out. Let your browser save it.
Set up video
The first boot downloads the models in the background. An amber banner at the top of the studio shows each download’s progress.
Video needs a free Hugging Face token before its models can download, so a Set up video dialog opens after a minute or so. Accept the terms on the two linked pages, create a read token, paste it in and click Save. (You can open the dialog again from the banner, or use the field under Settings → Models.)
Features unlock as their models arrive. Everything is usually ready in about 10 minutes, depending on the machine’s network. Until a feature’s models are ready, its Generate button is disabled. This only happens once: the models stay on your volume when you stop and restart the pod.
Make your first video
On the Create page, pick Video at the bottom, describe the shot and what you’d hear, and press Generate (or ⌘/Ctrl + Enter). The first clip takes a few minutes, because the video model loads the first time; later clips are quicker. For a still image, pick Image instead: it takes a few seconds.
