Studio

The Studio turns a prompt, and optionally a few files, into an image, a video clip, a sound, speech, or music. This page covers generating in the Studio, who pays for a generation, where results are saved, and how to generate from Files, chat, agents, and workflows.

What the Studio is

The Studio is the generation side of the home page. Use the Chat / Studio switch on the home page to move between the two. Each Studio thread is a conversation of its own: it appears in the conversation sidebar (filter it with the Studio chip) and keeps every generation you made in it. A thread is only created when you send your first request, so opening the Studio and leaving creates nothing.

The model you choose decides the format, the settings you can change, and the price. There are five formats:

Studio output formats
Format	Produces
Image	Still images.
Video	Video clips.
Sound	Sound effects and other audio.
Speech	A voice reading your text.
Music	Music tracks.

Generate an asset in the Studio

Choose a model
Open the model picker (Choose a model, or Change model once one is selected). Pick What to make (the format), then the Provider, then the model.
Choose who pays
Under the model list, set Credential source to Platform or My credential. See Who pays for a generation.
Describe it
Type your prompt in the composer (Describe what to create with the model you chose). Some models also run with no prompt, on their files and settings only.
Set the parameters and add files
Open Parameters to change the settings the model accepts. To give the model a picture or a clip, press Add a file and answer What is this file for?. Settings the model requires are highlighted until you fill them in.
Create
Check the price shown next to the button, then press Create. A video can take a few minutes. The result appears as a card in the thread.

Parameters

Each model lists its own parameters, so not every model shows every one. The common ones are:

Common generation parameters
Parameter	What it sets
What to avoid	A negative prompt: what should not appear.
Aspect ratio	The shape of an image or video frame.
Resolution	The output size.
Quality	The quality level the model offers.
Duration (seconds)	The length of a video or audio clip.
How many	The number of images to produce in one request.
Seed	A number that makes a result reproducible on models that support it.
Voice, Language, Style	Voice and style options, on models that have them.

When a parameter has a fixed list of values, the Studio shows them. Some lists are read from the provider when you open them (Asking the provider). If the list cannot be read, or is long and shown as a sample, you can still type a value.

Reference files

A model that accepts files asks what each file is for. The roles it offers depend on the model, and some roles cannot be combined (the Studio says which go together):

Roles of the files you attach
Role	What the model does with it
Source image	Transforms it: the file comes back changed.
First frame	The clip opens on this image.
Last frame	The clip ends on this image.
Reference image	Borrows its subject or style. It never appears as a frame.
Mask	Marks where the other image may be changed.

Reuse and modify a result

On a result card, Modify loads the prompt, model, settings, and files that made it back into the composer, so you can change one thing and create again. See it in Files opens the saved file. On an empty Studio, Your generations lists what your workspace has already generated, and the same actions are available there.

Who pays for a generation

Every generation runs on one of two keys:

Who pays for a generation
Choice	Who is billed	Price shown
Platform	You, in credits, at the rate published for that model.	Before you create: N credits for this request, computed from your settings (a longer video or a longer text costs more). After: Billed on the quantity measured.
My credential	The provider bills you directly, on a provider key you connected. The platform charges nothing for the generation itself.	No credit price is shown.

Platform is offered only for a model the platform actually sells: a platform key exists for that provider and a price is published for that model. Otherwise the model shows Not sold on the platform key and the Studio switches you to My credential, where you can connect a key for that provider without leaving the picker. Generations are measured per call, per second, per image, or per character, depending on the model.

Available models

The generation catalog is shared by the Studio, the Generate dialog, agents, and the Generate node. The catalog shipped with LiveContext includes models from AudioCraft, Clipdrop, Deepgram, ElevenLabs, Flux, Google Gemini, HeyGen, Higgsfield, Ideogram, OpenAI, Runway, Seedance, Stability AI, and xAI. What you can pick, and on which key, depends on what your install has enabled and priced. Generation models are not part of the chat model list in Settings › AI Providers (see Models & providers).

Where results are saved

Every generated asset is stored as a file in your workspace, so it outlives the provider's own temporary link. Open it with See it in Files. See Files.

Other ways to generate

From Files and chat

The Generate button in the Files toolbar, and the Generate an image, video or sound button in the chat tab of the side panel, open the Generate an asset dialog: Format, then Prompt (pick a model and describe it), then Result. The dialog uses the same models and prices as the Studio, and Past generations lets you start again from an earlier result. Files also lists your past generations beside the button.

From an agent or a chat

An agent can generate assets only when its Generation setting is on. It is off by default, for agents and for the chat assistant (in the chat settings), because every asset spends credits at the rate of the model used. See Agents and Chat.

In a workflow: the Generate node

The Generate node runs the same generation as a workflow step. Set the model, the prompt, the model's parameters, and who pays (platform, the default, or user for your own key). Whatever the format, the node always has the same outputs:

Generate node outputs
Output	Content
`file`	The generated asset as a file reference. Map the whole object into a downstream file parameter.
`model`	The model that ran.
`kind`	The format produced: image, video, audio, voice, or music.
`provider`	The provider whose model ran.
`billed_quantity`	The size the run was billed on (for example 10 for a 10-second video; 1 for a model sold per call).
`billed_unit`	What billed_quantity counts: call, second, image, or character.
`billed_credits`	Credits charged for the run. Absent when you paid the provider directly with your own key.
`provider_response`	The raw payload of the provider. Its shape varies by provider, so do not build on it.

Using the generated file downstream

{{agent:generate_cover.output.file}}

See Nodes for the other node types and Files for passing files between steps.

Troubleshooting

Studio messages and what they mean
What you see	What it means
No answer came back for this one.	The request may still have finished and been charged. Check your files before running it again.
This was refused before it ran, so nothing was charged.	Nothing was charged, and your prompt and files are kept. Check the settings, the files, or who pays, then create again.
It finished, but nothing came back that this page can show.	If an asset was produced, it is in your files.
This generation ran and was charged for, but the file could not be saved.	The generation ran and was charged, but saving failed. Use Save the asset from the provider right away: the link expires in a few minutes.
Not sold on the platform key	No platform price is published for this model. Use My credential with your own provider key.
Create images, videos and audio with a paid plan	Shown on the cloud when you press Create on the Free plan, which runs no generation model. Your prompt and files are kept: choose a paid plan, then create again. A workspace member sees This workspace’s plan does not include generation instead: ask the workspace owner to change plan.
Generation models require a paid plan. Rewards and top-ups do not unlock them.	A notice shown on a self-hosted install, which has no plans dialog, and in the Generate an asset dialog when no generation model is listed at all. Generation models need a paid plan on the account that runs them: its owner must move to a paid plan.

FilesWhere generated assets are stored and how to pass them on.NodesThe Generate node and the other workflow steps.Models & providersChat and agent models, keys, and bridges.