AI playable builder for mobile-game UA teams

AI playable ads in minutes, not weeks.

Agencies charge $5,000 to $15,000 per playable and take three weeks per round. By the time the variant lands, the meta you briefed against has already moved. Hand us a gameplay video, ship the variant set this week.

Apply for accessSee how it works
✓ AI build agent✓ Variant testing✓ MRAID export
SCORE0
×1
+0
TAP TO DROP
AI patch: "make explosion bigger"
variant_b - IPM 38.4
Winner detected
Variant C
↑ 47%
Time to ship
11 min
vs. 21 days
push to AppLovin
The playable production loop, inside one workspace
01 Game assets in
02 AI playable build
03 Plain-English edits
04 Missing assets generated
05 A/B variants rendered
06 Winner surfaced
07 MRAID ZIP export
08 AppLovin push
01 Game assets in
02 AI playable build
03 Plain-English edits
04 Missing assets generated
05 A/B variants rendered
06 Winner surfaced
07 MRAID ZIP export
08 AppLovin push
The pipeline is broken

You're being asked to find winners. You don't even ship the creative.

Three vendors, six revision rounds, and a five-figure invoice can land before the first impression. By then the format you briefed against may already be stale.

Today's pipelineUA team's calendar

Brief3 days
Concepting5 days
Build9 days
Revisions6 days
QA + ship2 days
7play~11 min
$8k to $15k
Per playable, agency rates
3 wks
Avg time-to-first-impression
UA managers are the ones being asked to find winners. But the build usually sits outside their hands, behind briefs, tickets, and vendor timelines.
7play moves that loop into the UA workspace. The team can test the mechanic, the hook, the CTA, the palette, and the audience angle before budget gets committed.
The result is not one precious creative. It is a set of variants with engagement feedback attached, so the next spend decision has evidence behind it.
Most current engines are template builders. You bring assets, pick a mechanic from a fixed gallery, swap in characters, export. 7play.ai is the simpler shape: hand us a gameplay video and the agents reason about everything else, including generating the sprites you do not have. You get playable variants grounded in your real game, not a skinned template.
How it works

If our marketing site can ship a 5-step playable, imagine what your studio can ship.

This section is itself a playable. Tap through the screens. It's the same edit loop you'll use to build yours.

STEP 01

Drop a gameplay video.

One input. No asset pack required. Optional: paste the store URL too.

+
drop assets here
Start building →
STEP 02

Agents study how your game plays.

Frame-by-frame analysis: mechanic, controls, visual language, UI, hook moments.

IPM 42.1
IPM 38.4
IPM 35.0
IPM 31.8
Start building →
STEP 03

Tell us what you want to test.

Hook, palette, CTA, mechanic variations. Short strategy chat. No tickets, no round-trips.

make the explosion bigger
✨patched fx_burst.scale 1.0 → 1.6+0.4s
move the CTA up
Start building →
STEP 04

Missing sprites? We generate them.

No isolated assets? Agents generate sprites in place against your game's visual language.

A
B
C
D
E
F
Start building →
STEP 05

Variants out. MRAID 2.0 to every network.

AppLovin, IronSource, Mintegral, Meta, Unity, Liftoff. Per-variant engagement flows back. The winner surfaces itself.

7
winner_v3
A
AppLovin
Status● Live
Impressions2.4M
IPM47.2 ↑
Start building →
01 / 05

Drop a gameplay video.

One input. No asset pack required. Optional: paste the store URL too. The agents have everything they need to get started.

Surgical editing

Direct your build like you'd direct a designer.

Tell us what you want changed. Every prompt becomes a scoped code, asset, or timing patch. You see the diff, preview the playable, and roll back anything that doesn't improve the variant.

  • ✓Surgical patches. Only the part you asked for changes.
  • ✓Undo edits, fork baselines, and branch variants without a handoff.
  • ✓Generate missing assets and audio inline. No more asset round-trips.
  • ✓Co-pilot suggests the edits viral playables in your category are using right now.
gem-quest · build #14
YOU
make the explosion bigger and move the CTA up
12:41 · sent
7P
scanning fx_burst.scale, ui.cta.y
Patched fx_burst.scale 1.0 → 1.6 and moved cta.y from 92% → 78%. Live preview ready.
+0.4s · 2 nodes · revert
YOU
add a level-2 fail state where the gem cracks
12:42 · sent
7P
generating crack_fx, wiring fail_state.l2
FAQ

Questions UA leads ask us.

Do I need to know how to code?+
No. The whole workspace is built around natural-language editing. You describe what you want changed; we patch the build. There's a JSON view for power users who want to peek under the hood, but you'll never need it.
How does the variant engine actually work?+
You pick the dimensions (hook, character pose, palette, CTA copy, etc.) and the values you want to test on each. We render every combination (the math is just the cartesian product) and serve them as a creative set. Engagement data per variant flows back in.
Which DSPs and ad networks do you support?+
Direct push to AppLovin, IronSource, and Mintegral. Universal MRAID 2.0 ZIP export works with any other DSP that accepts playable ads (Meta, Unity, Liftoff, etc.).
What about my existing assets and brand kit?+
Drop your characters, logos, screenshots, and UI kit into the reference pack. We auto-tag everything and use it as the source of truth. Every template skin and every variant is forced to stay on-brand. You set authenticity to "true-to-game" and it stays there.
Can I generate assets if I'm missing them?+
Yes. Character poses, FX, audio stings, and missing UI states can all be generated inline. Each generation has a fixed cost charged against your session budget cap, and you approve before it spends.
How fast is "minutes, not weeks"?+
The target workflow is a first playable preview in one session, then a multi-variant creative set after the first few edits. The exact time depends on asset readiness, brand constraints, and how many variants you want to test.
What happens if my brand team needs to approve creative?+
Every session has a shareable read-only preview link with comments. Brand can review variants, leave notes per-frame, and approve before anything is pushed live. Approval state is tracked per variant.
How is 7play.ai different from template-based playable engines?+
Honest answer: most current engines are template-based. You bring assets, pick a mechanic from a fixed gallery, swap in characters and copy, export. That ships reskins of known mechanics fast. 7play.ai is simpler at the surface: hand us a gameplay video, the agents reason about how your game plays (mechanic, controls, visual language, UI), figure out which assets matter, and generate sprites in place if you do not have isolated ones. Template builders scale creative. 7play.ai is creative discovery grounded in your real game.
What does it cost?+
We're in design partner mode right now, working closely with a small number of studios on their launches and creative scaling. Pricing follows the conversation, not the other way around. Drop us a line and we'll figure out what makes sense for you.
● Closed beta · design partners

Build 7play with us.

We’re shaping the product with a small group of UA teams and studios. Hand us a gameplay video and the agents build playable variants from your real game, no template gallery. If you ship playables now, or pay an agency five figures to, we want you in the room before we open the doors.

What design partners get

  • 01
    Full access during the beta

    Your games, your assets, real campaigns. Build a playable from a gameplay video and push the winner to AppLovin.

  • 02
    A direct line to the team

    Weekly calls with the founders. What you hit, and what breaks, sets what we build next.

  • 03
    Founding-partner pricing

    When paid plans open, you keep the founding rate. Pricing follows the conversation, not a price sheet.

  • 04
    Integrations we build for you

    We prioritize the ad networks and tools your team already runs on.

What we ask in return

  • ✓Active use on at least one live title or campaign
  • ✓A 30 minute feedback call every couple of weeks
  • ✓Honest reactions, rough edges included
  • ✓An optional case study once you ship a winning variant
Application · closed betaLimited seats this cohort

Tell us about your team.

How many playables do you ship per month?
We read every application and reply within 5 business days.