Hail Caesar is a mobile-first, full-screen AI comic strategy game for the Gemini Live Agent Challenge.
You are emperor. Every turn:
- Gemini writes the new state of Rome.
- Gemini generates 3 new comic panels (top-left scene 2, top-right scene 3, full-width bottom scene 1).
- Narration reads the turn with comic bubble highlights.
- Consequence and next actions appear.
- Your choice or spoken decree branches the campaign.
This build targets Creative Storyteller.
It uses interleaved mixed output per turn:
narrative_textvisual_description_for_imagenstat_changesplayer_options
See buildInterleavedBundle in server.js.
- Gemini SDK (required):
@google/genai - Text model:
gemini-2.5-flash - Image model (Nano Banana):
gemini-2.5-flash-image - Frontend: vanilla web app + Web Speech API (
SpeechRecognition,speechSynthesis) - Backend: Node.js + Express
- Google Cloud service: Cloud Run (deployment script included)
- Leverages a Gemini model: Yes
- Built with Google GenAI SDK or ADK: Yes (
@google/genai) - Uses at least one Google Cloud service: Yes (Cloud Run deployment path)
- Multimodal input/output: Yes (voice input, narrated audio output, generated visuals, structured interleaved turn data)
cd /Users/edidiong/Desktop/hackerthon/aura-narrator-live
npm install
cp .env.example .env
npm startOpen http://127.0.0.1:8787
gcloud auth login
gcloud config set project YOUR_GCP_PROJECT_ID
gcloud services enable run.googleapis.com cloudbuild.googleapis.com artifactregistry.googleapis.comchmod +x scripts/deploy-cloud-run.sh
./scripts/deploy-cloud-run.sh YOUR_GCP_PROJECT_ID us-central1 hail-caesarThe script builds and deploys the backend to Cloud Run, then prints the service URL.
flowchart LR
U["Mobile Player"] --> FE["Web UI (Comic Panels + Voice Controls)"]
FE --> API["Express Backend (Cloud Run)"]
API --> GEMTXT["Gemini 2.5 Flash via @google/genai (story turn JSON)"]
API --> GEMIMG["Gemini 2.5 Flash Image (Nano Banana panels)"]
API --> FE
FE --> U
- Public repo URL
<4 mindemo video showing real turn progression- short Cloud Console recording showing Cloud Run service running
- architecture diagram (above)
PORTdefault8787GEMINI_API_KEY(Gemini Developer API mode)GEMINI_MODELdefaultgemini-2.5-flashIMAGE_MODELdefaultgemini-2.5-flash-imageIMAGE_ENABLEDdefaulttrueTEXT_TIMEOUT_MSdefault26000IMAGE_TIMEOUT_MSdefault50000GOOGLE_GENAI_USE_VERTEXAIsettrueto use Vertex AI modeGOOGLE_CLOUD_PROJECTrequired whenGOOGLE_GENAI_USE_VERTEXAI=trueGOOGLE_CLOUD_LOCATIONdefaultus-central1