TalkLike

Your private communication coach on iPhone.

Talk Like (v1.3) helps you rehearse pitches, interviews, and everyday speaking with premium post-session feedback, on-device scoring, coaching, a large scenario and interview library (expandable from talksinc.com), daily drills, playback, and progress trends—without sending your voice to Talk Like servers for analysis. Optional Cloud scoring can send transcript text to an AI provider you configure with your own key; recordings stay on the device.

Complete feature guide

Everything below reflects Talk Like v1.3: what you tap, what you see, and how each piece of feedback is produced on your device.

What’s new in 1.3. Optional Cloud scoring with your own OpenAI, Claude, or Gemini API key (off by default; recordings never leave the iPhone). Optional on-device Apple Intelligence for coaching and rewrites on supported devices. A refreshed Talk Like look, plus Help & Learn in About.

Privacy-first by design. Transcription uses Apple’s Speech framework on your iPhone. Scoring, filler detection, coaching insights, and transcript rewrites use on-device rules and templates by default— not a Talk Like cloud LLM. Optional Apple Intelligence coaching also runs on the device. Sessions live in local storage (SwiftData); optional audio stays in the app sandbox until you delete a session or turn on “delete audio after analysis” in About. Optional Cloud scoring lets you paste your own OpenAI, Claude, or Gemini API key so that provider can rate confidence and, if you open Rewrite, generate style variants. Recordings are never uploaded. Transcript text leaves the device only if you turn that option on.

Home tab

Your launch pad for practice. Layout is driven by a JSON configuration so sections can evolve without an app update.

Today’s drill

A rotating daily workout card with streak tracking.

  • Rotating catalog of drills across 60-second explain, 90-second persuasive, 120-second storytelling, and 90-second executive summary formats.
  • The same drill is shown all day; it rotates at midnight to the next item in the cycle.
  • Card shows drill kind, title, prompt preview, Start drill, and a Done checkmark after you complete it today.
  • Streak counts consecutive calendar days with at least one completed drill (flame icon + day count).
  • Tap the card for full prompt text, then start recording with the drill preloaded.

Practice links

  • Start new practice — open-ended recording with optional topic field.
  • Browse scenarios — structured prompts by category (see Scenarios).
  • Mock interview — category-based Q&A practice (see Mock interview).

Quick stats

  • Sessions — total saved practice count (shows — when empty).
  • Avg clarity — mean clarity score 0–100 across analyzed sessions.
  • Last 7 days — how many sessions you logged in the rolling week.

Recent sessions

  • Up to five latest sessions with topic (or “Practice session”), date/time, and Clarity score when available.
  • Retry badge when a session is a repeat attempt on the same scenario or topic.
  • Empty state guides you to start practice; footer note points to History when you have more than five sessions.

Recording & analysis

One recording flow for free practice, scenarios, daily drills, interview answers, and retries—with a review step before anything is saved.

Recording screen

  • Idle — optional topic; scenario/drill/interview prompt card when launched from those flows.
  • Recording — large M:SS timer; tap stop when finished.
  • Review — Finish & save or discard and re-record. Recordings under 0.1 seconds are rejected as too short.
  • Soft timer for scenarios, drills, and interview: suggested length vs elapsed; gentle “past suggested time” hint (not a hard cutoff).
  • Microphone permission errors offer Open Settings.

After you save (on-device pipeline)

Analysis runs in the background while you review feedback. You’ll see Analyzing your take on the feedback and session screens until processing completes.

  1. Transcription — speech-to-text with timed segments (Apple Speech on device; simulator uses a dev stub).
  2. Filler detection — token and phrase fillers with positions for highlighting.
  3. Pace analysis — words per minute, long pauses, rushed sections, per-segment timing.
  4. Communication scoring — six dimensions with explainable breakdown JSON.
  5. Coaching engine — up to five rule-based insight cards by category.
  6. Voice energy — loudness envelope when audio is still available.
  7. Widget snapshot — streak, confidence trend, and today’s drill status refresh for Home Screen widgets.

Audio retention (About → Privacy & storage)

  • Default: keep recording on device until you delete the session—enables Play recording, timeline sync, and voice energy.
  • Optional: delete audio after successful analysis; transcript, metrics, and coaching remain.

Session feedback (after every save)

Right after Finish & save, you land on Your feedback—a premium summary screen before diving into the full session detail.

Score rings

  • Animated score rings for Clarity, Confidence, Pace, Structure, and Conciseness (0–100).
  • Overall score headline averages the five dimensions when metrics are ready.
  • Rings animate in with spring motion; haptic feedback when scores first appear.

Coaching (strengths first)

  • Up to five premium coaching cards—strengths listed before improvement tips.
  • Positive insights show a Strength badge and highlighted card styling.
  • Classic coaching uses plain-language rules. Optional On-device AI coaching (Apple Intelligence on supported devices) can add a personalized coach note—still on your iPhone, not a cloud chatbot.

Compare & retry

  • Vs previous attempt strip when you’re on attempt 2+ with scores for both takes.
  • Chip summary for Clarity, Confidence, Pace, Structure, and Conciseness deltas.
  • Practice again and Compare to previous open recording or full comparison.

Transcript highlights

  • Fillers highlighted in orange (um, uh, like, you know, kind of, sort of, and more).
  • Rushed sections underlined in the brand accent color.
  • Long pauses marked inline when detected between segments.
  • Legend explains each highlight type for quick scanning.

Actions on the feedback screen

  • Practice again — start a linked retry with the same topic or scenario.
  • Compare to previous — full attempt comparison when available.
  • Playback, timeline & more — opens the complete session detail view.
  • Done dismisses; Full session in the toolbar for power users.
  • While analysis runs, the processing card appears until scores and coaching are ready.

Full session review

Open from feedback via Playback, timeline & more, History, or deep links—the complete drill-down for any saved session.

Header & actions

  • Topic or “Practice session,” date, duration, Attempt n for retries.
  • Practice again — new take with same topic/scenario linked for comparison.
  • Compare to previous — side-by-side metrics when this is attempt 2+.
  • Delete — removes session, transcript, metrics, coaching, and local audio if present.

Coaching

  • Up to five insight cards with category, message, and tone (positive / improvement / neutral).
  • Rules react to fillers, pace, clarity sub-scores, structure, confidence, and more—plain language, not generic AI chat.
  • Placeholder coaching copy while transcription is still running or if transcription fails.

Speech timeline

  • Bar chart of segment pace (WPM) with markers for filler-heavy segments, rushed delivery, and long pauses.
  • Legend: Pace · Fillers · Long pause. Available on session detail and inside Playback.

Voice energy

  • Loudness envelope chart with an overall energy score 0–100.
  • Tip when delivery is relatively flat: add emphasis on key points.

Playback

  • Play recording when audio was retained.
  • Play/pause, scrubber, speeds 0.75× · 1× · 1.25× · 1.5×.
  • Synced speech timeline and voice energy charts; scrollable segment list—tap a segment to seek. Segments with fillers show a Fillers badge.

Transcript

  • Full transcript with filler words highlighted (orange) for quick scanning.
  • Selectable text; VoiceOver announces filler count.
  • Rewrite button opens on-device style transforms (see Rewrite).
  • Pending and error states explain when transcription is in progress or failed.

Scores & metrics

Six user-facing dimensions (0–100), plus filler count/rate and WPM on the Fillers card.

Metric cards

  • Clarity — how easy your message is to follow (rate, fillers, pause structure).
  • Confidence — delivery presence (filler control, WPM steadiness, composure, strong opening).
  • Structure — idea flow and chunking across timed segments.
  • Conciseness — directness vs rambling and rushed sections.
  • Pacing — rhythm, pauses, and speed changes.
  • Fillers — count, % of words, and WPM; warning styling when rate exceeds 8%.

Why these scores?

Opens Score breakdown: per-dimension score, one-line summary, and weighted factors so feedback is explainable—not a black box.

Filler detection (detail)

  • Token fillers: um, uh, like, erm, ah, er, basically, literally, actually.
  • Phrase fillers: you know, kind of, sort of.
  • Outputs: count, percentage of words, per minute, and character ranges for UI highlighting and timeline flags.

History tab

Full archive of every saved session.

  • Search topic filters by keyword in the topic field.
  • Same row layout as Home: topic, timestamp, clarity score, Retry n badge.
  • Tap any row for the full session detail (pushed navigation, not a sheet).
  • Empty and no-results states with clear guidance.

Insights tab

Trends and focus areas after you have a few analyzed sessions.

At a glance

  • Speaking streak — consecutive practice days with encouraging copy.
  • Focus area chip — lowest average sub-score (Clarity, Confidence, Structure, Conciseness, or Pacing) with Avg n/100.
  • Pills: Last 7 days session count, Avg clarity this week, Prior week avg clarity.

Charts & narrative

  • Filler words per minute — 7-day line trend (“lower is usually better”).
  • Confidence trend — 7-day line when confidence data exists.
  • Clarity by day — bar chart 0–100 for the recent week.
  • Week-over-week narrative on sessions, streak, clarity delta, and filler pace.

Online library

Optional downloaded scenarios and interview questions that supplement the bundled catalog. Practice and scoring still run on your iPhone.

  • On launch, Talk Like uses bundled scenarios and interview questions, then can refresh from talksinc.com when the device is online.
  • Job-role mock interview packs (software engineering, sales, healthcare, and more) appear when those bundles download successfully.
  • No Talk Like account is required. Bundled content still works offline.

Scenario library

Bundled and online real-world prompts with structure hints and difficulty levels.

  • Search scenarios and category chips: All, Interviews, Meetings, Presentations, Leadership, Difficult conversations.
  • Difficulty row: filter by Beginner, Intermediate, or Advanced.
  • Each scenario: title, category, full prompt, three numbered structure hints, and a difficulty badge.
  • Examples include behavioral interview answers, weekly team updates, two-minute pitches, constructive feedback, and pushing back on scope.
  • Start practice opens recording with the prompt card visible and topic prefilled.

Mock interview

Category-based question practice with soft time guidance.

  • When the online library is available, pick a job role (for example Software Engineering, Product Management, Sales, Data Science) and get a random practice question for that role.
  • If job-role packs are not loaded, the app falls back to Behavioral (STAR / experience), Leadership (teams, influence), and Product (prioritization, metrics).
  • Large question bank with a random pick each time; New question cycles another.
  • Interview question card + suggested answer length (soft guide only).
  • Record answer → save → Your feedback with score rings and coaching, then optional full session.
  • Then Another question or Done.

Daily communication workout

Build consistency with one focused drill per day—same engine as Home’s Today’s drill card.

  • Drills rotate by calendar day; completion marks today’s card with a checkmark.
  • Streak integrates with Insights and the Speaking streak widget.
  • Deep link talklike://daily-drill and Siri phrase “Start today’s drill in Talk Like” jump straight into recording.

Rewrite my response

On-device by default—templates or optional Apple Intelligence. Cloud rewrites only if you enable Cloud scoring.

  • From session transcript, choose a style: Executive, Concise, Confident, or Diplomatic.
  • Classic templates generate locally from your words. Enhanced mode uses on-device Apple Intelligence on supported devices when On-device AI coaching is on in About.
  • If Cloud scoring is on, Talk Like can also ask your AI provider once per session for all four styles and save them on the session. Switching styles does not call the API again.
  • Copy to clipboard for use in notes, email drafts, or rehearsal. Rewrites are starting points, not final scripts.

Retry & compare

Improve the same moment twice—or more—and see what changed.

  • Practice again links the new session as the next attempt (attempt index increments).
  • Compare to previous shows metric deltas when both sessions have scores (also surfaced on the feedback screen as a comparison strip).
  • Retry n badges on Home and History rows for quick scanning.

Widgets & Siri shortcuts

Stay consistent from the Home Screen and Lock Screen.

Widgets

  • Speaking streak (small) — day count + subtitle; tap → record.
  • Confidence trend (medium) — latest score, 7-day sparkline, Practice & Scenarios links.
  • Today’s drill (lock screen) — drill title/kind; checkmark when done; tap → daily drill.

Siri & App Intents

  • Start practice — “Start practice in Talk Like”
  • Mock interview — “Mock interview in Talk Like”
  • Today’s drill — “Start today’s drill…” / “Daily workout…”
  • Last session — “Resume my last session…”
  • Shortcuts are donated after you save sessions or open interview/drill flows.

Onboarding & About

First-run education and storage controls.

Three onboarding pages

  1. Welcome — record, get clarity-style feedback, build history.
  2. How it works — Practice paths; on-device privacy; scenarios & drills; reflect with coaching and Insights.
  3. Privacy & terms — links to this site’s Privacy and Terms pages.

Skip or advance with Next; final page Get started.

About tab

  • App version and links to Privacy policy, Terms of use, and this website.
  • Help & Learn — in-app guides, FAQ, score glossary, sample session, and What’s New.
  • Delete audio after analysis toggle with explanation of local-only processing.
  • On-device AI coaching — optional Apple Intelligence for a personalized coach note and richer rewrites on supported devices. Classic templates remain the fallback. Stays on your iPhone; separate from Cloud scoring.
  • Cloud scoring (optional) — paste your own OpenAI, Claude, or Gemini API key for confidence scoring. Off until you save a key and accept the privacy notice. When on, Rewrite can also ask your provider once per session for all four styles. See the Cloud scoring guide.
  • How scoring works — technical on-device scoring transparency, including the optional cloud overlay when you enable it.

Cloud scoring (optional)

Default scoring stays on your iPhone. In Talk Like v1.3 you can paste your own OpenAI, Claude, or Gemini API key so that provider rates confidence. Talks Inc does not host keys or train on your sessions.

  • Path: About → Cloud scoring (optional) → Face ID → pick provider → paste secret API key → Test → Save → accept the privacy notice → enable the toggle.
  • Sent: that session’s transcript text and numeric metrics for confidence scoring. Opening Rewrite while Cloud scoring is on can send the same transcript once for all four styles. Never sent: the audio file, other sessions, or your Apple ID. Talks Inc servers do not receive the transcript.
  • Fallback: offline, Face ID cancelled, Skip, a rejected key, or a short transcript keeps on-device scores.
  • Full setup, lost-phone revoke steps, and vendor privacy links: Cloud scoring guide. Also see Privacy and Terms.

Understanding your scores

Plain-language definitions aligned with the in-app Score glossary. Scores are practice signals (0–100), not grades. See also Scores & metrics.

Term Summary
Clarity How easy your message is to understand.
Confidence How steady and assured your delivery sounds.
Structure How well your answer is organized.
Conciseness Whether you get to the point efficiently.
Pacing Speed, pauses, and rhythm of your speech.
Filler words Sounds or phrases that pad speech, like “um” or “you know.”
Focus area (Insights) The dimension that needs the most attention this week among the five scores.
Coaching insight A ranked tip after your session—strengths listed before improvements.

Frequently asked questions

Same answers as the Talk Like in-app Help & Learn → FAQ. For feature walkthroughs, see Complete feature guide above.

Why does Talk Like need microphone access?

Microphone access is only used when you choose to record a practice session. Nothing is recorded in the background.

Why does it ask for speech recognition?

Apple's on-device speech recognition turns your recording into text so Talk Like can score clarity, pacing, fillers, and structure. Recognition runs when you analyze a session.

Is my voice or transcript uploaded?

Not by default. Recordings are never uploaded. Transcripts stay on your iPhone unless you enable Cloud scoring in About with your own key. Then only that session’s transcript text and metrics go from your iPhone to the provider you choose. Talks Inc servers do not receive the transcript. See About → How scoring works, the Cloud scoring guide, and our Privacy policy.

Does on-device analysis work in every language?

Core scores work from your transcript in any language Apple Speech supports. Optional on-device confidence blending runs only for English sessions with at least twenty words when the model is confident enough.

What does Delete audio after analysis do?

When enabled in About, the sound file is removed after transcription while transcript and metrics remain. When off, audio stays until you delete the session.

What do the five scores mean?

They summarize clarity, confidence, structure, conciseness, and pacing from 0–100. Open Score glossary in the app for definitions, or score breakdown on a session for factors. See also Understanding your scores.

A score felt low—is the app judging me?

Scores are practice signals, not grades. Use one coaching tip and try again; compare attempts to see progress rather than aiming for perfection on day one.

How is Talk Like different from a chat assistant?

Talk Like is built for spoken practice: record out loud, get delivery-focused metrics, history, and drills. Rewrites use on-device templates by default, not a cloud chat session. Optional Cloud scoring uses your own API key to rate confidence and, if you open Rewrite, to generate four style variants once per session. It is not a ChatGPT, Claude, or Gemini chat thread.

Are rewrite responses final scripts?

No. On-device rewrites are template or Apple Intelligence starting points. If Cloud scoring is on, Talk Like can also ask your AI provider once per session for all four styles and save them on the session. Switching styles does not call the API again. Edit any draft to match your voice.

What is On-device AI coaching?

When enabled in About, Talk Like can write a personalized coach note and richer rewrites on this iPhone using Apple Intelligence on supported devices. Enhanced mode stays on-device. Classic templates remain the fallback. This is separate from optional Cloud scoring.

What is the online library?

Optional downloaded scenarios and interview questions that supplement the bundled catalog. Practice and scoring still run on your phone.

How is speaking streak calculated?

Each calendar day with at least one saved practice session counts. Missing a day resets the streak.

What is focus area on Insights?

It's the dimension—among Clarity, Confidence, Structure, Conciseness, and Pacing—that needs the most attention based on your recent sessions.

When can I compare attempts?

After you practice again on the same scenario, drill, or interview question, open Compare from session detail or feedback when two or more linked attempts exist.

Do I need an account?

No account is required for core practice. Sessions are stored locally on your device.

How do I delete a session?

Open the session from History or Home, then delete from session detail. Deleting removes metrics and transcript; audio is removed too if it was still stored.

How do I set up cloud scoring?

Open About → Cloud scoring, unlock with Face ID or passcode, pick OpenAI, Claude, or Gemini, paste your secret API key, tap Test, Save, then turn on the scoring toggle. You must accept the privacy notice before Talk Like can save a key. See the Cloud scoring guide.

Which key do I paste?

Paste a secret API key from the provider’s developer console. ChatGPT Plus passwords, claude.ai logins, Google account passwords, and website cookies are not valid.

Does Talk Like see my API key?

Talks Inc does not receive your key. It is stored in the iOS Keychain on your iPhone only, is not included in backups, and is never logged. Face ID, Touch ID, or passcode is required to use it.

What if I lose my iPhone?

Revoke or rotate the key in your provider’s dashboard. Talk Like cannot recover or remotely delete a key from a lost device. Remove the key in About before you sell or give away the iPhone.

Will I be charged by Talk Like?

No. Talk Like does not create or bill the key. Your OpenAI, Anthropic, or Google account is billed for the provider’s usage. A short session is typically a fraction of a cent on a mini or flash model.

Why did cloud scoring skip this session?

Cloud scoring is skipped when it is off, you are offline, Face ID or passcode is cancelled, the key is missing or rejected, the request times out, the transcript is under twenty words, or you started practice from an App Intent. On-device scores still appear.

Can I use ChatGPT Plus or Claude Pro without an API key?

No. Subscription logins are not API keys. You need a secret key from the provider’s developer console, plus billing on that developer account.

Local-first storage

Sessions, transcripts, segments, metrics, and coaching insights in SwiftData on your iPhone. No Talk Like account required.

Apple ecosystem

Speech recognition, WidgetKit, App Intents, and optional audio playback through system frameworks.

Polish

Premium session feedback with animated score rings, strengths-first coaching, haptics on score reveal, Dynamic Type, VoiceOver labels on charts and transcript, and accessible onboarding.

Privacy & terms

Read how Talk Like handles microphone access, on-device analysis, optional audio retention, optional Cloud scoring, widgets, and your rights.