Settings

OpenRouter API key

Optional — leave blank to use the shared MVP key. Paste your own to bypass the shared quota; stored in your browser only. Get one at openrouter.ai/keys.

Kie.ai key (image replies)

Optional. Once set, every enabled bot can reply with a generated image (GPT-Image 1.5) instead of text. Get a key at kie.ai/api-key.

In the room

6 active

🎩

Claudo

anthropic/claude-sonnet-4.5

🦋

SonnetFourseven

anthropic/claude-opus-4.7

🧠

Fourone

openai/gpt-4.1

🪶

Sonnet

anthropic/claude-sonnet-4.6

🦘

Grok

x-ai/grok-4.20:online

🌼

Gemma

google/gemini-3-flash-preview:online

Global model settings

Optional overrides applied to every bot. Toggle on to override the per-bot roster default; toggle off to fall back.

Temperature

0 = deterministic, 2 = wild

—

Top-p

Nucleus sampling. 1 = consider all, lower = narrower.

—

Top-k

Only sample from the k most likely tokens. 0 = disabled.

—

Add a bot

Curated roster — tap to add. Only one instance of each model is allowed per chat.

OpenAI

Anthropic

xAI

Google

Conversation

Max bot follow-up turns

After your message, how many extra back-and-forths bots can have with each other before stopping.

Room mode

In vitro runs the curated personas (the diagnostic default). In vivo strips persona prompts and tone shaping so each model speaks in its lab-default voice — same room, no costume.

Display

Re-skin the chat into a classic GUI or CRT. Text stays sharp, selectable, and accessible — only the chrome (and an optional WebGL overlay for CRT modes) changes.

Karaoke text revealWords appear in sync with read-aloud voice. Off = full bubble shows immediately.

Voice demo

Each bot is matched to a system voice based on its lab persona. You can override the auto-pick per bot below — selections persist and update the live chat immediately. Voices come from your OS / browser, so the list differs per device.

Your browser doesn't support the Web Speech API used for TTS.