Settings
Add download tokens, check your models, and download setup logs.
Open Settings from the sidebar (on a phone, from the tab bar at the bottom). Most changes save when you click out of a field.
Download tokens
- Civitai token: needed to import some LoRAs from Civitai on the LoRAs page, including NSFW ones.
- Hugging Face token: raises Hugging Face download limits, and is needed for gated models (ones whose terms you must accept on Hugging Face first), including LTX-2.5, the studio’s default video engine.
Models
Shows each group of models your pod downloads and whether it’s Ready, Downloading or Not enabled. Features that need a group unlock when it’s ready.
If a download is waiting on a gated model, a Hugging Face token field appears right under it. Paste a token there and click Save. The download starts within about 15 seconds, with no restart.
Groups that are Not enabled were switched off when you deployed. To add one, set its DOWNLOAD_*_MODELS variable to true with Edit Pod in Runpod (see Choose which models download); it downloads on the next start.
LTX-2.5, the default video engine, comes with sound and its own community license (free under $10M annual revenue, with a few restrictions — see the license for details). MiniMax H3 is an alternative you can enable instead; its community license excludes the US, EU, UK and South Korea unless you get a license from MiniMax.
Setup and logs
Shows how long each step of setup took on this pod — studio ready, video model ready, first video, and so on — timed from when the pod started. Useful for judging how long a fresh pod takes to get going.
Download logs saves a file with the studio’s logs, model and GPU status, and this setup timeline. It has no passwords, keys or other secrets, so it’s safe to attach to a bug report.
Sign out
The Sign out button is at the top of the Settings page. Sessions otherwise last 30 days.
Set keys when you deploy
Set these as environment variables under Set overrides on Runpod’s deploy page, or later with Edit Pod; a pod restart picks up the new value. ANTHROPIC_API_KEY and the OPENAI_* variables connect an AI for script breakdown and Enhance prompt — there’s no field for them on the Settings page. CIVITAI_TOKEN and HF_TOKEN can also be set from Download tokens above; if you set one in both places, the Settings page wins. STUDIO_PASSWORD works a little differently: when it’s set, it replaces the password you created in the studio.
| Variable | What it does |
|---|---|
STUDIO_PASSWORD |
Your studio password. Overrides the password you created on first visit. The template’s default, change-me, counts as not set |
ANTHROPIC_API_KEY |
Connects Anthropic for script breakdown and Enhance prompt |
OPENAI_API_KEY, OPENAI_BASE_URL, OPENAI_MODEL |
Connects an OpenAI-compatible provider instead |
CIVITAI_TOKEN |
Civitai token for LoRA imports |
HF_TOKEN |
Hugging Face token |
Choose which models download
Everything is on by default. To make the install smaller, set a switch to false under Set overrides before you deploy (or later with Edit Pod; a switch turned back on downloads on the next start).
| Variable | Models | About |
|---|---|---|
DOWNLOAD_IMAGE_MODELS |
Images (Z-Image Turbo) | 21 GB |
DOWNLOAD_EDIT_MODELS |
Edits, camera angles, storyboard frames (Qwen-Image-Edit) | 31 GB |
DOWNLOAD_LTX_MODELS |
Video with sound (LTX-2.5) | 40 GB |
DOWNLOAD_LTX_IC_MODELS |
Shots that follow a previs (LTX add-on) | 1 GB |
DOWNLOAD_LTX_INGREDIENTS_MODELS |
Reference sheets for consistent characters and props (LTX add-on) | 1 GB |
DOWNLOAD_VOICE_MODELS |
Character voices (Qwen3-TTS) | 9 GB |
DOWNLOAD_PERFORM_MODELS |
Perform (Wan Animate) | 25 GB |
DOWNLOAD_MINIMAX_MODELS, DOWNLOAD_MINIMAX_REF_MODELS |
MiniMax H3, an optional video engine (off by default; see its license above) | 42 GB each |
DOWNLOAD_UPSCALE_MODELS |
4K export upscale (SeedVR2, off by default) | 9 GB |
Agent access
An AI agent such as Claude Code can make films in your studio on its own: write the storyboard, render, review each shot and export. It signs in with an agent token instead of your password.
The Agent access section of Settings has the steps to connect one, including a command you can copy. Use an AI agent walks through it, including the previs skill. New token replaces the token, and agents using the old one stop working. If you set the token yourself with the STUDIO_AGENT_TOKEN variable, change it in Runpod instead.