A proposal to design the next chapter of your AI voice health & fitness app: the flows, the hi‑fi screens for iOS and Android, a Figma system that translates to Flutter without argument, prototypes your developers can build from, and the store screenshots that sell it. Six weeks, one small team, no surprises in week three.
It runs on Flutter. It has a Figma library, a logo and a body of product content. What it does not yet have is a design practice that keeps pace with the roadmap, and that is the gap this dossier closes.
Your posting asks for a product and UI designer to deliver five things. We have listed them below exactly as we read them, because we intend to be held to them: every deliverable has a chapter in this document and a week in the plan.
The brief's own phrases, which we will keep using because they are good ones:
A microphone icon in the corner is not voice-first. The test we will hold every screen to: can a user finish the three core jobs (log a meal, run a set, check progress) without looking? And can they finish the same three without speaking? Both answers have to be yes, or the kitchen and the train each lose a user.1
"Via voice, text or photo" is the most important clause in the brief. If those three paths produce three different records, you are maintaining three products. We design one entry shape and three ways to fill it, with correction per item rather than "say it again".
Kitchens are loud, gyms are louder, the 07:52 is quiet but public. So every spoken cue is mirrored as a caption, every voice path has an equal tap path, and the app never assumes it was heard correctly. Misheard and offline are first-class states, not error toasts.2
Screens age; a token-based system that maps one-to-one onto Flutter theme extensions is what keeps the next twenty screens consistent when we are no longer in the room. That is why chapter five is a diagram and not a mood board.
Voice is the front door, not a feature hidden behind one.
Three principles carry the whole system. Everything in chapters four and five is one of them made concrete.
The orb is the home screen. Idle, listening, thinking, speaking, misheard, offline: six states, one visual grammar, each mirrored in a caption so the screen says what the speaker says.
Voice, photo and text write the same entry. Each item carries its own confidence, and correction is per item: tap the one line that is wrong, never repeat the whole meal.
Rep counts and spoken cues live on the workout screen as captions. Insights arrive weekly, led by one plain sentence. The paywall comes after value is shown, with a seven-day trial and App Review-compliant terms.
"Two poached eggs, sourdough toast and a flat white."
| Poached eggs ×2 | 156 kcal · 13 g |
| Sourdough toast 1 slice | 140 kcal · 5 g |
| Flat white semi-skimmed? | 120 kcal · 8 g |
| Breakfast | 416 kcal · 26 g protein |
Same card, whichever door the user came through. The orange dot marks the one item the app is least sure of, so the correction is one tap on one line. Values illustrative.
Tap the steps, press play, or use the arrow keys. The phone is live: the orb, the waveform and the transcript are the same state machine the Flutter app will implement.
Nothing to read, one thing to do. The orb sits still, the mic is the largest target on screen, and the two smaller buttons (camera, menu) are the equal tap paths. The caption says exactly how to begin, so the screen never relies on the user knowing a magic phrase.
Design tokens leave Figma as JSON. Your Flutter team consumes that JSON as a theme extension. Nobody retypes a hex value, and the voice state machine is documented once and shared by the design file, the prototype and the app.
Each week ends with something you can open, use and argue with. Weeks one, three and five are decision gates: we do not move on until you have signed the thing off, and we say so in writing on the Friday.
| Week | Milestone | What you receive | How we know it is done |
|---|---|---|---|
| 1Gate | Discovery & auditWe read before we draw. | Audit of the existing Figma library and app builds; inventory of current flows; token gap list; the three-job test written up as a script. | You sign off the flow map and the gap list. Figures in chapter seven are confirmed or revised here. |
| 2 | Flows & wireframesRoadmap features, low fidelity. | User flows and wireframes for voice mode, unified logging, live workout, insights, onboarding and paywall; the voice state machine as a diagram with timings. | Two rounds of Figma comments closed; every screen has a tap path and a voice path. |
| 3Gate | System foundationsVariables, primitives, the contract. | Figma variables in light and dark modes; primitive and pattern components with variants; first tokens.json export; Flutter mapping document. | One of your Flutter developers builds one screen from tokens alone, without asking us a colour question. |
| 4 | Hi-fi screens IThe voice-first core. | Voice mode (all states), logging by voice, photo and text, live workout with captions. iOS and Android variants from the same components. | Three-job test passed on device by two people new to the file. |
| 5Gate | Hi-fi screens II & prototypesInsights, onboarding, paywall. | Weekly insights, onboarding, subscription paywall (7-day trial, monthly and annual, en-GB). Interactive prototypes with developer annotations. | You approve the prototypes as the build reference; the handoff doc is read by your team, not just received. |
| 6 | Store assets & handoffShip-ready. | App Store and Google Play screenshots in every required size; UI kit documentation; a recorded 20-minute walkthrough; 30 days of async support after handoff. | Your team can ship a new screen without us in the room. |
The base covers every deliverable in your posting. The options below are the things clients ask for in week four; tick them now and the total updates, or leave them and add them later at the same price. Figures are in pounds, exclusive of VAT, and marked illustrative because we have not yet seen your file.
No hourly meter. If a week runs long because we misjudged it, that is ours. If it runs long because scope moved, we say so on the Friday and price the change before we do it.
Same team, same cadence, rolling monthly with two weeks' notice. Suits a roadmap that will keep producing features after week six. Illustrative.
Copywriting beyond UI strings, nutrition database licensing, App Store and Play developer accounts, and Flutter implementation itself, which we can quote separately once the system exists.
Exaflair is a product-engineering firm. Our designers sit next to Flutter engineers, which is why chapter five is a contract, not a wish. Handoff is a habit here, not a phase.
We have already published four invented-brand explorations of voice-first coaching, including the alternatives we set aside. This dossier is the fifth. You are not buying a promise to think; the thinking is on the table.
Around twenty-four screens is a small number for this scope. That is deliberate: a voice-first app wins by removing screens, and a design system wins when nobody needs a new one.
We bring the audit checklist and the three-job test script; you bring the Figma file on a shared screen. By the end of the call we will both know whether the week-one figures hold, and the clock can start the following Monday.
Prepared with care, and to be argued with. — Exaflair, September 2026