TL; DR
- The opening was rebuilt. The first act is now a hero’s journey
- Backend is off localhost: live at our web host, full pipeline in 265 seconds from a phone
- Budget submitted: $7,040, not to exceed
- Quarters Presentation, Sep 11.
· · ·
The Opening, Rebuilt
Two pieces of feedback landed last week, and both were right.
| The problem | What we did |
|---|---|
| “The two-layer world is too hard to grasp.” | One world. A near future where living virtually is simply normal. No lore, no slogan, no “Aidol” label. Teleports, one-tap outfit morphs, floating panels, and nobody reacts. |
| “The opening has no hero’s-journey shape.” | The first act now has turns. Ordinary day → call → refusal → mentor → threshold → audition. |
The new first act, following Vogler’s stages:
| Stage | Length | What happens |
|---|---|---|
| Ordinary World | 30–40s | An ordinary day in the virtual city |
| Call to Adventure | 15–20s | Echo rushes in, begging for a stand-in |
| Refusal – Agree | 20–30s | “Why me?” — Echo persuades the player |
| Crossing the Threshold | 5–10s | The player says yes, and walks on stage |
| Tests – Ordeal – Reward | ~2–3 min | Audition → judges’ verdict → PASS card |
Ordinary World
Call to Adventure
Refusal – Agree
Crossing the Threshold
Tests – Ordeal – Reward
· · ·
Our Current Build
Experience:
Using a pre-defined look, it records your own voice and then alters your voice to a pre-defined voice, then generate an idol’s self-intro video.
Time & Cost:
About five minutes per run. Around $1.30 of real vendor spend, each time, on us. We also have a ledger (via Supabase) that records every call: request, response, latency, billable units, cost in USD.
· · ·
Cost Breakdown
| Per session | Cost | What’s in the chain |
|---|---|---|
| Today | ~$1.30 | No image generation yet, cheaper video model for testing |
| End of V1 | ~$4.73 | Per-player image generation in, Seedance 2.0 at the quality bar |
| Planning cap | $10.00 | Headroom, so a rate move gets absorbed rather than escalated |
Budget approved for all AI services but fal.ai and Liblib AI, so we switch to Pollo.ai (approved).
| Service | Type | Purpose | Amount | Date |
|---|---|---|---|---|
| Pollo.ai | Subscription | Video generation | Waiting for reimbursement | Sep 3 |
| LiblibAI | Subscription | Video generation | CNY 199 | Sep 7 |
| Fal.ai | API credit | Image and video generation | $21.40 | Sep 8 |
$21.40 USD + CNY 199
· · ·
Weekly Progress Report
Meeting Notes:
Feedback
- Split budget documentation in two: API tokens, and subscriptions that buy team efficiency. Two different arguments, two different sections.
- “This should not be all free.” Players need to understand from early on that things in this world cost something.
- Tie the minigames to the narrative. The matching game currently feels disconnected from the story around it.
- Add a PR person in-game. Pay them to get booked on a show, and you buy fan reach — relationship as an economic system.
Still blocked
- The startup application path through CTTEC / Swartz Center.
- Reimbursement approval for Pollo.ai, ElevenLabs, Suno and Topaz. These are pipeline stages, and they don’t run without it.
· · ·
Quarters Presentation