How to get an API key for the AI generator

The AI generator runs on your own Anthropic or OpenAI account. You need an API key from one of them, and the account behind the key needs prepaid credit. Setting it up takes about five minutes. Innerscene charges nothing.

What you need

An account on the provider's developer platform: platform.claude.com for Anthropic, platform.openai.com for OpenAI.
Prepaid credit on that account, at least $5. A key on an account with no credit is the most common reason the first model fails.
A payment card. Claude Max, Ultra and Team plans include monthly API credit, which you can use with no card (step 3 under Anthropic).
A ChatGPT Plus or Claude Pro subscription does not cover API use. The API is billed separately.

Which provider to pick

Either works. Claude Opus 5.5 from Anthropic is the default and cost less per model in our tests. If you already have an OpenAI developer account with credit, use that. You can save a key for each provider and switch between them in the panel.

Get a key from Anthropic (Claude)

Open the Claude Console at platform.claude.com and sign up or sign in. The old address, console.anthropic.com, leads there too. The Console is Anthropic's developer account: you can sign in with the email you use for the Claude app, and it has its own billing.
If the Console asks about your organization and how you will use the API, fill it in. An individual can enter their own name.
Open Billing (Settings > Billing), press Buy credits and buy at least $5. Credits are ready to use straight away. With a Claude Max, Ultra or Team plan you can claim the plan's monthly API credit: on claude.ai open Settings > Billing, find API credits and choose Link organization (how it works).
Optional: on the same Billing page, under Spend limits, press Set limit and enter a monthly maximum, for example $20.
Open API keys (Settings > API keys) and press Create key. Give it a name such as "Innerscene viewer".
In the same form, pick a workspace for the key (the default workspace is fine). The panel needs a key that belongs to one workspace. For Expiration, pick a length that suits you; when the key expires you will need a new one. Leave Linked account set to yourself.
Copy the key. It starts with sk-ant-, and the Console shows it only once. If you lose it, create a new one.
In the AI panel choose Use your own AI key, open the Anthropic (Claude) tab, paste the key and press Check and save key.

Both Claude models in the panel, Claude Opus 5.5 and Claude Sonnet 5.5, are available on every Anthropic usage tier. A new account can start with lower rate limits for its first days, so an early request may be asked to wait.

Get a key from OpenAI (GPT)

Open the OpenAI platform at platform.openai.com and sign up or sign in. Your ChatGPT login works here.
Open Billing (Settings > Organization > Billing), add a payment method and buy at least $5 of credit. Your first $5 moves the account to the Build usage tier, which both OpenAI models in the panel need.
Optional: open Limits (Settings > Organization > Limits), under Spend choose Edit spend limit, enter a monthly amount and turn on Enforce a hard limit.
Open API keys and press Create new secret key. Give it a name, keep the default project and leave Permissions at All. A restricted key must at least be allowed to list models and to use the Responses API. OpenAI may ask you to confirm a phone number first.
Copy the key. It starts with sk- (usually sk-proj-) and is shown only once.
In the AI panel choose Use your own AI key, open the OpenAI (GPT) tab, paste the key and press Check and save key.

OpenAI lists GPT-6 Astra and GPT-6.1 Sol from the Build tier upward, so buy credit before you start. OpenAI asks some organizations to complete a Verified Organization ID check before they can use certain models; if the panel shows a message about verification, see Organization verification.

What the key check does

Check and save key asks the provider for its list of models. This costs nothing and proves the key is genuine. It does not show whether the account has credit, so a key can pass the check and the first model can still fail with the no-credit message.

What a model costs

The provider takes the cost of each request from your credit. These are estimates from our own tests with the default models, Claude Opus 5.5 and GPT-6 Astra; your bill can differ.

What you make	Standard quality	Detailed quality
3D object or jewelry	$0.05 to $0.30	$0.30 to $1.00
Person	$0.05 to $0.30	$0.20 to $0.80
Floor plan (DWG)	$0.05 to $0.75	$0.30 to $1.50

Each follow-up message costs about the same again.
GPT-6 Astra costs more per token than Claude Opus 5.5, so OpenAI results tend toward the top of each range. Claude Sonnet 5.5 and GPT-6.1 Sol cost less and suit simple objects.
The panel shows a running estimate for the session. The provider's bill is the final figure.
$5 of credit covers roughly 20 to 50 models at Standard quality.

List prices per million tokens

Model	Provider	Input	Output
Claude Opus 5.5	Anthropic	$4	$20
Claude Sonnet 5.5	Anthropic	$2	$10
GPT-6 Astra	OpenAI	$10	$50
GPT-6.1 Sol	OpenAI	$2	$10

Prices change. Check Anthropic pricing and OpenAI pricing.

Keeping your key safe

Stored only in this browser. The panel saves your key in this browser's local storage for innerscene.com, on this device. Other browsers and devices never have it. If your browser blocks site storage, as some private windows do, the key lasts only until you close the page.
Sent only to the provider. Your browser sends the key straight to api.anthropic.com or api.openai.com with each request. It never goes to Innerscene's servers and is never part of a report or feedback you send.
What Innerscene receives. Anonymous usage statistics: the steps you use, the model and quality chosen, token counts, the cost estimate and whether a model was made. No key, no prompt text and no replies. More about privacy
Innerscene's own account. Innerscene staff signed in to the company's systems can use the company account. Their requests go through Innerscene's server and use no personal key. Other visitors do not see this option.
Use a dedicated key with a spend limit. Create a key just for this site and set a monthly spend limit, so a leaked or mistaken key can only cost what you allowed.

Remove the key from this browser

In the panel, press Forget key. Clearing this site's data in your browser settings removes it too. On a shared computer, do this when you finish.

Revoke or replace the key at the provider

Anthropic: on the API keys page, Disable stops a key and can be undone; Delete removes it for good.
OpenAI: on the API keys page, delete the key to revoke it.
To replace a key, create a new one, paste it in the panel with Change key, then delete the old one at the provider.

Troubleshooting

The panel shows these messages in English. Find yours below.

The key is rejected

The panel says: The provider rejected this API key. Check it, or create a new one.

Check that you copied the whole key with no spaces, and that it is on the matching tab: Anthropic keys start with sk-ant-.
The key may have been deleted, disabled or expired. Create a new one.
OpenAI also refuses requests from countries it does not serve (list), and requests from a restricted key that may not list models.

No credit on the account

The panel says: Your account with the provider has no credit left. Add credit in their billing settings, then try again.

Buy credit on Anthropic Billing or OpenAI Billing, then try again.
Claude Max, Ultra and Team subscribers: link the Console organization to the plan (Anthropic step 3) so the monthly credit arrives.

Too many requests

The panel says: The provider is limiting requests right now. Wait a minute and try again.

Wait a minute and send the message again. New accounts start with lower limits that rise with use.
If waiting does not help, the account has probably reached a monthly limit: an Anthropic usage tier cap (see Billing), or an OpenAI spend limit you set (see Limits). Raise the limit or wait for the next month.

The model is not available

The panel says: This model is not available to your API key. Pick another model.

Pick the provider's other model in the panel settings.
OpenAI: buy at least $5 of credit so the account reaches the Build tier, or complete organization verification.

The provider is busy

The panel says: The provider is busy or did not answer in time. Try again.

The panel already retries a few times. Wait a minute and try again. Anthropic calls this state "overloaded".
Check Anthropic status or OpenAI status.

The provider cannot be reached

The panel says: Could not reach the provider. Check your connection and try again.

Your browser talks to api.anthropic.com or api.openai.com directly. Ad blockers, privacy extensions, a company firewall or proxy, or a VPN can block these addresses.
Allow both addresses, pause the extension for innerscene.com, or try another network or browser.

The key does not belong to a workspace (Anthropic)

The provider's message starts with: This API key is not scoped to a workspace…

Create a new key on API keys and pick a workspace in the form (Anthropic step 6). Then paste the new key with Change key.

Your own spend limit was reached (Anthropic)

The provider's message starts with: You have reached your specified API usage limits…

Raise or remove the limit on Billing, or wait until it resets.

Organization verification (OpenAI)

If a message says your organization must be verified, open Organization settings and start the verification there. OpenAI checks a government ID, and access can take a few minutes to arrive afterwards.
Until then, pick the other OpenAI model or use an Anthropic key.

The panel asks for the key every time

Your browser is not keeping site storage, which happens in private windows and with some privacy settings. Allow site data for innerscene.com, or paste the key on each visit.

No key? Join the credits waiting list

Innerscene is building prepaid credits, so you can use the generator with no key of your own. Open the panel, choose Buy credits from Innerscene and press Join the beta list. We will email you when credits are ready.

Open the AI generator

Official pages

Checked on 1 November 2026. The providers change their pages from time to time; theirs are the final word.

Back to the AI generator · Full guide to the AI generator: refining, exporting and privacy