How Does SLIDD AI Work? The 3-Stage Reply Engine
How Does SLIDD AI Work? Inside the 3-Stage Reply Engine
So how does SLIDD AI work? In three stages that happen in under a second: it captures your screen with iOS's ReplayKit the moment you tap Reply, a vision model reads what's on it, and a second model writes your reply in your chosen tone before dropping it straight into the text field. Nothing gets uploaded anywhere, and nothing gets saved afterward.
That's the whole pipeline — one architecture doing in a single tap what screenshot-based competitors split across eight separate steps and two different apps.
How Does SLIDD AI Work When You Tap Reply?
Tapping Reply on SLIDD triggers a three-stage pipeline: ReplayKit captures the active screen, a vision-capable model reads the message, photo, or profile on it, and a language model drafts a reply in your selected tone before inserting it directly into the text field — the entire sequence finishes before you'd finish reading this sentence.
You're not switching apps to make this happen. You're switching keyboards, the same way you'd switch to an emoji keyboard, except this one can see.
The three stages, in order:
- Capture — the visible screen is read the instant you tap Reply
- Understand — a vision model interprets what's actually on it
- Write — a second model produces the reply and inserts it
Each stage runs automatically. You pick the tone once, tap once, and the reply lands in the field ready to send.
Key Takeaways
- SLIDD reads your screen in real time using iOS's ReplayKit, capturing exactly what she sent and the surrounding context without you leaving the dating app.
- The pipeline runs in three stages — screen capture, vision-model interpretation, reply generation — and finishes in under a second.
- SLIDD works inside any app with a keyboard (Hinge, Tinder, Bumble, Instagram, WhatsApp, iMessage) because it's a system keyboard, not a separate dating app.
- Five tone presets plus a custom tone builder let the reply match the conversation's energy instead of sounding generated.
- Screen content is processed and discarded immediately — nothing is stored, nothing trains a model, and nothing is tied to your identity.
How Does SLIDD AI Read Your Screen in Real Time?
SLIDD uses Apple's ReplayKit framework — the same screen-broadcast technology built for screen recording — to capture the active app frame the moment you tap Reply. It doesn't record continuously in the background; the capture happens once, on demand, when you ask for a reply.
This is the part every competitor skips. What SLIDD AI actually is comes down to this single design choice: the keyboard doesn't need you to hand it information, because it can already see the screen you're looking at.
There's no permission dance beyond the one-time iOS screen-broadcast prompt, and no separate camera roll access. ReplayKit hands the frame to SLIDD's processing pipeline, and that frame — her message, her photo, her Story, her bio — becomes the raw material for the next stage.
How Does the Vision Model Turn a Screen Capture Into a Reply?
A vision-capable model reads the captured frame the way you would — it picks out the message text, notices a photo if one's attached, and registers where the conversation sits (opener, mid-thread, or the moment right before you ask her out) — then a second model writes the reply in your chosen tone and drops it straight into the text field.
It isn't reusing a stock line; it's generating fresh text calibrated to what the vision model just saw. The finished reply gets inserted directly into whatever text field is active — no preview screen, no copy button, no paste step.
Worth knowing: SLIDD's stage detection is what keeps replies from sounding uniform — an opener gets written differently than a reply sent five minutes before a date gets locked in.
Is Your Data Safe When SLIDD Reads Your Screen?
Screen content is processed the moment you tap Reply and discarded immediately afterward — SLIDD doesn't store conversations, doesn't use them to train models, and doesn't tie anything it reads back to your identity. The capture exists only long enough to generate your reply.
That's a meaningfully different privacy model than a tool that asks you to upload a screenshot to a server. There, the image has to travel somewhere and sit in a queue before anything happens to it. SLIDD's read-and-discard approach skips that step because the processing happens close to the moment of capture, not after a round trip.
Why Screenshot-Based Competitors Can't Match This Architecture
Rizz AI, Keys AI, YourMove AI, and RizzGPT all share the same broken sequence: screenshot her message, leave the dating app, open the other app, upload the image, wait, copy the reply, switch back, paste it, send. That's eight steps standing between her message and anything you actually type back.
Same goal, different number of steps — and the gap is architectural, not cosmetic.
Rebuilding around real-time screen reading isn't a feature update for those tools — it means abandoning the screenshot-upload model their entire product is built on. RizzGPT still asks users to manually paste conversation text in. YourMove AI and Keys AI both route through a separate app before a reply exists. None of them read the screen the way SLIDD does, which is why SLIDD's architecture keeps beating Wing AI and every other screenshot-based tool on the one metric that actually matters mid-conversation: speed from her message to your reply.
Reality check: Eight steps sounds manageable in the abstract. At 11pm, mid-conversation, with her typing back, it's usually enough friction that the reply never gets sent at all.
Which Tone Should You Choose?
The same pipeline that reads the screen also decides which tone fits it best.
SLIDD offers five tone presets — Flirty, Bold, Witty, Sincere, and Casual — plus a custom tone builder, because the same reply logic shouldn't answer "how was your day" and a late-night message the same way. Each tone changes how the language model writes the reply, not just its word choice.
Tap Witty and SLIDD turns "lol I bet you say that to all the girls" into "Only the ones who almost catch me lying." Tap Sincere on the same message and it writes something closer to "Honestly? You're the exception, not the pattern." Same input, different register — that's the whole point of separating tones.
Casual and Sincere are available on the free tier; Flirty, Bold, and Witty unlock with a paid plan, along with the custom tone builder. SLIDD's full pricing and feature breakdown covers exactly what's included at each tier, but every new signup starts with three days of unlimited access to all five tones before any of that matters.
How Does SLIDD AI Work Across Every App You Actually Use?
Yes — SLIDD works anywhere iOS lets you switch keyboards, including Hinge, Tinder, Bumble, Instagram DMs and Stories, WhatsApp, iMessage, Snapchat, Telegram, and Discord, because it's a system keyboard rather than a standalone dating app. Competitors built as dedicated apps stop working the moment a conversation moves off the dating platform.
That matters more than it sounds like it should. A match that starts on Hinge usually moves to Instagram, then WhatsApp, then iMessage — and every jump is a place a screenshot-based tool goes silent. SLIDD doesn't notice the jump because the keyboard is already installed everywhere text gets typed.
Get Started with SLIDD AI
Now that you know exactly how the reply gets written, the only step left is turning it on. Every new account gets three days of unlimited access to every tone before you decide on a plan.
FAQ
Does SLIDD store or save the conversations it reads? No. SLIDD processes screen content in real time and discards it immediately after generating a reply. Conversations aren't stored on SLIDD's servers, aren't used to train any model, and aren't linked back to your identity — the same standard iOS screen-broadcast permission used for screen recording is what makes the read possible.
Is SLIDD AI free to use? Every signup gets a 3-day unlimited free trial with every tone unlocked and no payment required upfront. After that, plans run $6.99/week, $19.99/month, or $99.99/year, all of which unlock unlimited replies, all five tones, the custom tone builder, and full reply history.
How fast does SLIDD actually generate a reply? The full pipeline — screen capture, vision-model read, reply generation, and insertion into the text field — completes in about a second. There's no upload step and no processing queue, since the read and the write happen back-to-back on the same tap.
Does SLIDD work on Android? Not yet. SLIDD is iOS-only at launch, live in the US and Canada App Store. Android support is on the roadmap but isn't available currently, so Android users can't install the keyboard extension yet.