Product

How Scaler works

Short answer: Scaler runs a closed loop — generate → publish → capture the outcome → label → retrain — with no human curation gate in the middle. You brief it once on your product. After that it detects what is working in your niche, produces content, publishes it through official platform APIs, and records whether each post produced a paying user.

The distinguishing property is the label it learns against. Most tools optimise for views because views are the metric platforms hand you for free. Scaler optimises for conversions, which requires connecting a revenue source and is the reason the loop has to close.

Stage 1 — Brief once

You describe the product, who it is for, and how it should sound. That brief becomes persistent memory the agent grounds every later decision against, rather than context you re-supply each time. Concretely it holds things like category, audience, tonal constraints, and claims the agent is allowed to make about your product — so generated content stays factually tied to what you actually shipped.

Stage 2 — Detect what is moving

The agent scans your niche for formats, hooks and sounds gaining velocity, not ones that already peaked. The output is not a trend report — it is a testable hypothesis: this format, with this hook structure, for this audience, should outperform the current baseline. Hypotheses are what the next stage is built to falsify.

Stage 3 — Generate, with the variable isolated

Script, storyboard and asset are produced together. The important design decision here: the hook or overlay text varies independently of the underlying visual asset. If both changed at once you could never tell whether a win came from the message or the creative. Separating them is what makes the results interpretable, and it is the mechanism the whole feedback loop rests on.

Stage 4 — Publish through official APIs

Content publishes to TikTok, YouTube Shorts and Instagram Reels using your own OAuth authorization, through the platforms' official publishing surfaces — TikTok's Content Posting API and the YouTube Data API. No device spoofing, no unofficial automation, nothing that puts the account at risk.

One constraint worth knowing: YouTube accepts video only. Image-based formats such as carousels and slideshows publish to TikTok and Instagram, and are assembled into video when they need a YouTube surface.

Stage 5 — Attribute, label, retrain

Each post gets a unique tracked link at publish time. Conversions come back through a connected Stripe or RevenueCat account, so every post carries its own view → install → revenue chain rather than a view count alone.

That conversion outcome is then labelled and fed back as the training signal for the next generation round. This is the part that compounds: the tenth week of output is informed by nine weeks of labelled results about what converted for your product specifically. The measurement caveats are real and worth understanding — we wrote them up in detail.

What runs without you

What still needs you: the initial brief, connecting your billing account, and the judgement call about whether a claim the agent wants to make about your product is actually true.

Frequently asked

Does it post without my approval?

By default yes — running without a human curation gate is the design intent. You can review drafts before publish if you prefer, but the system is built to operate without that step.

Which platforms are supported?

TikTok, YouTube Shorts and Instagram Reels. YouTube is video-only, so image formats route to TikTok and Instagram.

Is this against TikTok's terms?

No. Publishing goes through the official Content Posting API with your own authorization — the same surface any approved developer uses.

What does it optimise for?

Paying users, not views. That requires a connected Stripe or RevenueCat account; without one, Scaler can report engagement but not conversion.

See pricing →