Playground
Try chat, image, video, music and speech models from the dashboard without writing code, billed and logged like API calls.
The playground is part of the dashboard at /app/playground. Sign in, pick a tool, and run models from your browser. Each tool sends the same requests your code would send to the API, so what you see is what you get from code.
Billing and logs#
Playground calls run on a hidden key that belongs to your account. You don't create it and it isn't listed under API keys. Every call is charged to your balance at the model's normal price, and appears in Logs next to your API traffic, under the key name Playground. See Usage & logs and Models & pricing for how charges are computed; live prices are on /models.
Tools#
Talk to any chat model, with its speed and cost under every reply.
Generate and edit pictures, with your saved characters.
Cast characters and places, write shots with @tags and dialogue, join them into one video.
Put your character into any clip, keeping its motion.
A video that keeps going while you steer it.
A song in, a music video out, with the singer lip-syncing every line.
Turn text into speech with preset voices, voices you describe, or your own voice.
The people and places you saved, for every image and video.
You only see the tools your account has models for.
Find the playground#
- On a computer, open the Playground group in the sidebar. Each tool is a link in it; Characters & scenes is the last link, below a divider. Click the group name to fold it away; the dashboard remembers your choice.
- On a phone, tap Playground in the tab bar at the bottom. It opens a page with one card per tool and a Characters & scenes card below them.
Every tool has a Docs link next to its title that opens the matching API docs.
From playground to code#
Most tools have a View code button in the header. It shows a cURL command for the request you are about to send, with your current model and settings filled in and $LOVELY_API_KEY in place of the key. Create a key in API keys, set it as LOVELY_API_KEY, and run the command. The Quickstart covers the first request.
Asleep, waking up, warm#
Some models run on our own GPUs (Hosted models). They go to sleep after about 2 minutes without use, and the next request waits while they start. When you pick one, a badge under the model picker shows its state:
| Badge | Meaning |
|---|---|
| Asleep | Your first request waits while it starts (the badge shows the usual range); after that it is fast. |
| Waking up | Starting up. You can keep typing meanwhile. |
| Warm | Ready: it answers right away. |
Picking a model or looking at the badge doesn't wake it: only a request you send does. For video and music models the badge can't show a state, so it says how much longer the first run after a quiet spell can take.