The six streams, one at a time
Each stream below has its own calculator with its own inputs. Set them to whatever you believe is realistic and the monthly contribution updates. Nothing here is a projection — it is arithmetic on numbers you choose.
Combined monthly contribution
Sums every calculator belowTake these to unit economics to subtract infrastructure and see actual profit.
01 · Gems & live gifts
Largest streamFans buy gem packs at full price and spend them on gifts during live streams. A gift worth N gems creates N beans, the split engine divides the beans, and the creator redeems at 200 beans to the dollar.
The two levers that matter are the split percentage and the purchase channel. Moving a buyer from in-app purchase to web checkout is worth more than ten points of split.
Run bonus-gem promotions instead — buy 100, get 120. The fan pays full price and perceives more value, while the revenue per transaction is unchanged. Discounting the pack price compresses margin permanently because the bean redemption rate is fixed.
02 · Fan subscriptions
Most predictableRecurring monthly memberships to a creator's fan club with tiered perks — exclusive gifts, badges, private chat, monthly calls.
Worth more per dollar than gifting because it is predictable and it does not pass through the bean conversion. The platform's 20% is 20% of the actual price, not 20% of a pool.
At 8% monthly churn a subscriber lasts about 12 months. At 15% they last under 7. The lifetime value roughly halves, and no amount of pricing work recovers it. Retention features — exclusive gifts, member-only streams — pay back faster here than anywhere else in the model.
03 · Digital products & store
Highest marginCourses, templates, ebooks, community access, software and UGC items — sold by creators, delivered digitally, commissioned by the platform.
The best-margin stream in the model: no inventory, no shipping, no bean conversion, and the commission is charged on the full sale price. The digital_products table already tracks creator_earnings and platform_fee per purchase.
04 · Live commerce
Highest volume potentialPhysical goods sold live on stream with shoppable overlays, plus affiliate commission on third-party catalogues.
Large transacted volume, thin take rate. The risk is that returns and chargebacks are settled in physical goods, which the reserve fund was not designed to cover — model it separately from gifting.
Commission is earned on the order but usually reversed on a return. At a 20% return rate a 10% take rate behaves like 8%. Set the input below to what the category actually does, not to zero.
05 · Brand deals & sponsorships
LumpiestSponsored campaigns, kingdom-event sponsorships, and the brand-to-creator deal marketplace where the platform charges a fee on the transaction rather than selling the audience.
Few transactions, large values. A handful of deals can swing a month, which makes it a poor stream to forecast against and a good one to build on top of a base of gifting and subscriptions.
06 · B2B, SDK & AI twins
Only one with a real cost of goodsMini-game SDK licensing, the translation API, enterprise AI twins, and creator AI twins that charge per question, per month or per consulting session.
The ai_twins table already carries all three pricing modes — price_per_question, price_monthly and price_session — and ai_twin_interactions records the split per answer. What it does not do is record what the answer cost to generate.
Every other stream on this page has effectively zero marginal cost. An AI twin answer costs real tokens. The tokens_used column exists and nothing writes to it, so today there is no way to tell a profitable twin from one that is underwater. Set a hard token ceiling per interaction and a daily spend cap per user before this stream is switched on — the margin bar below shows why.