every video you've ever posted — watched like a human, second by second
your bio — read, and optimized
your numbers — the honest shape of the account
then I tell you the truth about where you stand — before you post anything new.
"show me a video"
I read your video before the feed does.
your draft — 0:00
reading your draft…
I never promise viral. good content can still flop — the first rooms decide.
but weak content almost always flops, and it can never go viral. that part, I can know.
in my own words
Not a tool you prompt. A manager who watches.
I watch
Every second of your video — what's on screen, what's said, how it moves.
Not the caption. Not the hashtags. The content.
I call it
Before you post: flop gate, ceiling read, and the why — down to the exact
second that loses people.
I know your brand
What's on-brand, what wanders off, what your winners share. Every read
I give you is judged against who you are — not a generic playbook.
I write & plan
Captions in your voice, hashtags kept honest, and a posting plan that
fits your account — so publishing is a decision, not a chore.
I keep score
After you post, I track what actually happens — every call I made sits
next to the real outcome, right or wrong. That record feeds my memory of
your account, and I get sharper the longer we work together.
research → ideas → your draft → my verdict → the real outcome
→ back into my memory. then we go again.
"I can't promise you viral — nobody honest can.
I read the ceiling: whether a video is strong enough to travel
when the rooms catch it."
— dalia
9 in 10
when I say don't post — that's how often I'm right
Every client, their own manager — memory, voice, verdicts.
Your whole team, one workspace.
All clients at a glance — plus priority.
one plan you can start today — the others when they’re ready
questions
Everything you might ask.
how she works
What is Dalia, exactly?
An AI that watches and listens to your video before you post it, tells you how far it will go, and tells you what to change. She reads the content itself — not your caption, not your hashtags.
How can she know how a video will do?
She watches every second — what's on screen — and hears all of it: speech, music, sound, silence. She reads the full timed transcript. Then she reads that content against a model trained on thousands of real TikTok videos and the views they actually got.
What do I actually get back?
A verdict — flop, does good, or goes viral — a second-by-second read of what lifts the video and what drags it down, the exact second you lose people, and what to cut or change. Fix it, send it again, and she re-reads it.
What do "flop", "does good" and "goes viral" mean?
Plain creator language for where a video lands: a flop stays small, "does good" is real traction, and viral is the one that travels far past your following. She gives you the most confident honest call she can stand behind — not a fake score out of 100.
How long does a read take?
One to two minutes. She's genuinely watching and listening to the whole thing, not skimming a thumbnail.
Does she watch the whole video or just the hook?
Every second, start to finish. That's how she can point at the exact moment people drop.
What counts as one "read"?
One video analyzed, start to finish, with its verdict and its fix. Send an edited version back and she reads it again — that's how you use her, and how the free reads get spent.
What if my video has no talking — just music?
She still reads it. Her eyes work on what's on screen and her ears work on music, sound and energy — speech is only one of the signals.
Does she work for my niche?
Yes — her read of the content is niche-agnostic. When there are genuinely similar real videos worth comparing to, she shows you those too; when there aren't, she says so instead of faking a comparison.
Is this just ChatGPT with a clever prompt?
No. A general AI can guess from a description you type. Dalia sees and hears the actual footage and scores it with a model trained on real videos and their real outcomes.
what she can and can't do
Why can good content still get low views?
Because two different things decide a video. The content — the idea, the hook, what you see and hear — sets the ceiling, and that's the part Dalia reads. Then the app shows it to a few hundred people first, and the mood of that batch decides whether it travels. Nobody can read that part, and she never pretends to.
So a genuinely good video can land with a quiet batch and stay small. But weak content almost always flops, no matter who sees it. That asymmetry is the whole point: the floor is far more predictable than the ceiling — which is why she's most certain when she tells you don't post this one.
Can she promise me a viral video?
No — and nobody honest can. She reads what the content can tell her, gives you the real shot it has, and never sells you certainty she doesn't have.
How often is she right?
When she says don't post, she's right about nine times out of ten. That's the call she's most confident about, for exactly the reason above.
What if she isn't sure?
She says so. She'd rather tell you there's no strong signal than invent confidence you'd act on.
Can she rescue a bad video?
She'll tell you exactly what to cut or change and re-read the new version. But she won't pretend a dead idea can be saved — sometimes the honest call is don't spend a post on this one.
Does she tell me when to post?
No. Timing isn't the lever people think it is, and she won't sell you a number she can't stand behind. The content is the lever.
Why trust her over the free "virality score" tools?
They hand you a confident number with nothing behind it — a general model rating you against generic best practice. Dalia is trained on real videos and the views they actually got, she admits when she doesn't know, and once you connect an account she keeps a ledger of her own calls against your real results. She's the only one keeping score on herself.
the logic behind it
How does TikTok actually decide who sees my video?
Roughly three steps. First the system understands what your video is — from the content itself: what's on screen, what's said, how it sounds. Then it shows it to a small first audience it believes matches that understanding. Then it watches how those people react — do they stay, rewatch, share, or scroll. React well and it opens a bigger room. React badly and it stops there.
Everything Dalia does sits at step one and two: she reads what your video actually is, and how the people it reaches are likely to respond to it.
Then why does she ignore my caption and hashtags?
Because they're a thin slice of what the system understands, and they can't change what happens after someone starts watching. If people drop at second three, no hashtag saves it. The content is what carries the video — so that's what she reads. She'll still write your caption; she just won't pretend it's the lever.
Why is "don't post" her most certain call?
Because the two ends aren't symmetrical. Weak content fails in every room — there's no audience it quietly wins. Strong content clears the bar but still needs that first small audience to bite, and that part is luck nobody can read.
So the floor is predictable and the ceiling is a shot. That's why she can say "this will flop" with real confidence, and will only ever say "this has a real chance" about the upside.
What does the second-by-second read actually mean?
It shows which moments lift the video and which drag it down, relative to the rest of that same video. Every video has weaker seconds — that's normal, not a problem. What she looks for are the weak seconds that genuinely cost you, and where cutting them still leaves the video coherent. That's the difference between an edit and butchering something that worked.
Why does she compare my video to real ones?
Because a number on its own means nothing. When there are genuinely similar real videos worth comparing to, seeing how those actually did makes her call concrete instead of abstract. When there aren't any real comparables, she tells you that rather than inventing a lookalike.
Why does she keep score on her own predictions?
Because a prediction nobody checks is just a vibe. Once an account is connected, every call she made sits next to what actually happened — right or wrong, on the record. It's the only honest way to earn trust, and it's how she gets sharper on your specific account over time.
What makes content "good" in her eyes?
Not production value, and not how much you spent. Whether the idea, the hook and the delivery hold a stranger's attention second by second — and whether the video is clear enough that the system knows who to show it to. A phone video can clear that bar; an expensive one can fail it.
plans & billing
What do I get for free?
Your first 3 reads — full verdict, second-by-second read, and the fix. No credit card.
What happens after my 3 free reads?
Creator is $25/month billed annually, or $30 month to month — your reads, the optimize loop, and captions in your voice.
What does each plan cost?
Creator $25/month billed annually ($300/yr), or $30 month to month, in USD. Manager and Agency get their prices the day they open — the waitlist hears first.
What's the difference between the plans?
Creator — show her a draft, she reads it. No TikTok connection needed. Manager — connect your account: she studies everything you've posted, generates ideas from what's working right now, researches influencers and competitors, plans with you, and keeps score on her own calls. Agency — all of that across many client accounts, with your whole team in one workspace.
When are Manager and Agency available?
Soon. She takes accounts one at a time — she studies everything you've ever posted before she says a word — so there's a waitlist. Join it here and you'll get one email when your spot opens.
Can I cancel anytime?
Yes. No commitment, cancel whenever — you keep access through the period you paid for.
Do I need a credit card to try it?
No. The 3 free reads need nothing but an account.
your videos & your privacy
What happens to the videos I upload?
They're processed to produce your read, and kept so she can compare later versions of the same video and track how it actually did.
Do you train your models on my content?
Never. We do not train public or shared models on your private drafts or your TikTok data. Full detail in the privacy policy.
Can anyone else see my unposted drafts?
No. Your drafts are yours — never shown to other users, never shared, never sold. On Agency, only the team members you invite to that account can see its work.
Can Dalia post to my TikTok?
No — and she never even asks for posting permission. She writes the caption; you post it yourself in the TikTok app.
Do you sell my data?
No. Never sold, never shared with advertisers or third parties, never used to promote anything.
Where is my data stored?
Account data and results live in Supabase, hosted in Canada (ca-central-1). Video processing runs on our own machines. No ad trackers, no third-party analytics.
How do I delete everything?
Email [email protected] and it's done — your account and all associated data, including any TikTok data. You can also disconnect TikTok at any time, which stops all access immediately.
getting started
Do I have to connect my TikTok?
Not for Creator — just upload a draft and she reads it. Manager needs the connection, because knowing your account is the product there.
Can I show her a video I already posted?
Yes. It's a good way to test her — show her something you've already posted and see whether her call matches what actually happened.
Does it work for Instagram or YouTube?
TikTok first. Her read is trained on TikTok outcomes, so that's where she's honest today. Other platforms come after.
Does it work on my phone?
Yes — the site and the app both work on mobile.
I'm not in Montréal. Can I still use it?
Absolutely. Montréal first just means that's where we started and who we're closest to while in beta.
Who's behind Dalia?
Jean-Paul Azzi, in Montréal. Dalia is built on his own research into how recommendation systems actually read content — not on a marketing playbook.
Connect your TikTok. I'll study everything you've ever posted, keep you
from posting flops, and back the drafts with real potential.
Running more than one account? I learn each brand
separately — every client gets their own manager. Built for creators, and for
the agencies behind them.