Oki Doki could not start.

Close and reopen the app, or reload this page. If it still does not start, update your browser or Android System WebView and try again.

WELCOME TO OKI DOKI ✿

A lovely place to begin.

Dating is fun, internet simulations are safe and happy and healthy.

Data tracking is ethical and cookies are a great idea.

Click here if you agree and to continue.

◉OKI DOKI THE LAST TICKET

Resume your saved story? Progress since that save, including an unsent message, will be lost.

THE LAST TICKET · A NIGHT WITH THE CREW

Oki
Doki

A rainy night, an arcade full of friends, and one last prize to win. The rules are simple. At least, they were.

Saving is optional and stores your choices and submitted messages on this device.

A short psychological mystery. Contains themes of loss and unsettling audio descriptions.

Location: TEAM HQTHE LAST TICKET / 01

When Continue is available, press Enter or Space to advance the story.

PERFECT MOMENT
A STORY IN PROGRESS
Recent story
    AN INDEPENDENT OPEN-SOURCE PROTOTYPEA PLAYABLE STORY ✳

    OPTIONAL · EXPERIMENTAL

    AI settings

    The story includes voices and is playable without an AI service. The desktop download includes an optional local character model and transcription. You can also connect Mirid or compatible services for dialogue and speech recognition.

    Use the included desktop AI

    The voices are already included. This sets up local character replies and microphone transcription; recording only starts when you choose it in the story. A custom speech service is kept.

    No account needed. Save the settings for this session after setup.

    Connect Mirid

    Checking lists advertised capabilities; it does not load a model or prove it is ready.

    Connect a phone to desktop Mirid

    Start the companion on the computer, then enter its trusted HTTPS address and pairing token.

    The companion uses its configured transcription engine and voice service.

    Character dialogue

    About AI instructions

    In many AI applications, users cannot see every system or developer instruction shaping a response; visibility depends on the service. Oki Doki’s instructions are in its open source. A connected provider may apply additional instructions of its own.

    For the desktop package with its local AI files. Starting the model uses memory and may take a moment.

    Model compatibility

    Choose the parameter your provider supports.

    Ignored when no output limit is selected. Some models count hidden reasoning tokens toward this limit; larger limits can increase cost and wait time.

    Choose provider default if your model does not accept a custom temperature.

    Slower local models may need more time. You can continue without waiting.

    Speech

    For microphone replies in the desktop package. Recording starts only when you choose it in the story.

    Keep voices on this browser for offline play (about 170 MiB). Each game update needs a new download.

    Faster preparation can reduce pauses between voices, but may send an extra request that isn't played. Your provider may charge for it.

    Realtime read-aloud requests a short-lived token from your paired companion when you select Read aloud or use Auto-read. The companion must be explicitly configured for OpenAI; generation may incur provider charges. Voice input continues to use the speech service below.

    Creator voice audition

    Audition your independently recorded reference for one speaker. This is a local session preview; it does not publish or change the voice for other players.

    Uploading sends your selected recording to the Chatterbox service URL above, where the service stores it as a voice reference. Nothing is uploaded until you select the button. The selected engine must be available in that service; the fixed-voice Linux Kokoro mode cannot use a reference.

    Select an engine and reference, then save the settings for this session.

    The experimental Linux voice normally uses Mirid’s fixed local voice and does not send the speech API key; the voice engine and voice fields below apply to the other services. A creator-assigned voice uses its own service.

    The default voices are Mirid Kokoro IDs. For an OpenAI-compatible service, enter a voice supported by your chosen provider in each field; OpenAI’s tts-1, for example, accepts alloy.

    Preview uses the settings currently shown here. It may generate speech locally or send a short sample to your selected speech service. The experimental fixed Linux service uses one voice. Realtime assigns distinct voices to the other speakers.

    Recording starts only when you select Record. Read aloud speaks one passage; Auto-read speaks each passage and its choices when shown until stopped. Speech can use included recordings or local generation; a configured speech service receives the text it speaks. Your audio, text and story context go to the services you configure. Later dialogue requests can include earlier private and group exchanges as context. Microphone permission is requested by your browser. Settings and keys stay in this page's memory and are cleared on reload or close.