Join the Echo Beta

Echo is built by one person around a full-time mail route — every tester genuinely moves the project. This page covers how the beta works, how to send feedback that actually helps, and structured test plans if you want to hunt bugs on purpose.

The short version: use Echo like you really listen — commute, chores, workouts, bedtime — and tell us the moment something feels wrong.

01Joining the beta

  1. Install TestFlight (Apple's free beta app) from the App Store.
  2. Open your Echo invite link on the device — it opens in TestFlight.
  3. Tap Accept, then Install. Echo appears on your home screen like any app, with an orange dot in TestFlight marking it as a beta.

Don't have an invite yet? Request access via GitHub Issues — include the device you'd test on.

Requirements: a recent iPhone (on-device alignment uses the Neural Engine, so newer hardware aligns faster). Apple Watch and Mac are optional but very welcome test surfaces — the watch app installs automatically from the iPhone's Watch app.

Updates: TestFlight notifies you when a new build lands; each build's What to Test notes tell you where to aim. Beta builds expire after 90 days; updating resets the clock.

Your data

Beta data carries forward to future builds — but it's a beta, so keep your source audiobook files backed up. (You should anyway; Echo never modifies them.)

02Sending feedback that helps

The fastest way: TestFlight screenshots

See something wrong? Take a screenshot right then. Tap the screenshot preview → ShareSend Beta Feedback. Your note arrives with the build number, device model, and iOS version attached automatically — that's half the diagnosis done.

Crashes

If Echo crashes, TestFlight offers to send the crash report with a comment box. Please add one sentence about what you were doing — a crash log with "happened when I tapped Auto-Align on a 40-hour book" is worth ten without.

Trackable bugs & ideas

For anything you want tracked (or to check whether it's known): github.com/dfakkeldy/Echo/issues. The developer reads every one.

Reporting alignment problems (special instructions)

Alignment bugs are the most valuable reports and need the most context. Include:

  • Book title + narrator (different narrations behave differently)
  • Where the ePub/PDF came from — store, edition, year if known (editions drift; that's often the whole bug)
  • What the audio is — single M4B, folder of MP3s, LibriVox, Libation export…
  • Where it went wrong — chapter, and roughly what the text said vs. what was playing
  • Whether Auto-Align had run, and whether you'd placed manual anchors
"Book X aligned perfectly except chapter 7 drifted ~30s after an ad-libbed intro; ePub is the 2nd edition from Kobo; audio is a Libation M4B" — that's a perfect report.

03Structured test plans

Pick whichever matches your life. Each is 10–20 minutes of intentional testing.

Plan A — The Commute Run (Smart Rewind & interruptions)

  1. Start a book, then live your interrupted life: pause for seconds, minutes, an hour.
  2. Each resume: did Smart Rewind back up a sensible amount? Ever dump you somewhere confusing?
  3. Mid-sentence, unplug headphones / disconnect Bluetooth. Echo should pause — never blast the speaker.
  4. Take a phone call. Afterward, does Echo resume if it was playing — and stay paused if you'd paused?

Plan B — The Alignment Gauntlet (EPUB + auto-align)

  1. Put an ePub in the audiobook's folder; confirm Echo imports it automatically.
  2. Run Auto-Align Chapters and watch the tiers run (plug in for long books — the first run also downloads the ~40 MB speech model).
  3. Spot-check five chapters: tap a paragraph — does audio land on those words, or close?
  4. Find the worst spot and fix it: long-press → Align to Now. Does surrounding text snap into place?
  5. Search a distinctive phrase; tap the result. Text and audio should jump together.

Plan C — The Study Session (bookmarks, flashcards, review)

  1. While listening, make three bookmarks: one plain, one with a voice memo, one with a photo.
  2. Re-listen across them: does the memo play inline? Does the artwork switch to your photo and back?
  3. Promote a bookmark to a flashcard (front as a question; attach the audio snippet).
  4. Tomorrow, when the review notification arrives: Daily Review on the phone — then a session on the watch, hands-free.
  5. Grade honestly; check the stats module updates (due / reviewed today / total).

Plan D — The Library Stress Test (formats & playlist)

  1. Load your messiest book: multi-file M4B, a 100-file LibriVox folder, weird filenames.
  2. Check the chapter list: grouping sensible? Sections under chapters where expected?
  3. Drag-reorder a few tracks; dim one (e.g., a disclaimer track) and confirm playback skips it.
  4. iCloud users: try a book without "Keep Downloaded" on cellular — how does Echo cope? Then enable it and compare.

Plan E — The Wrist-Only Day (watch remote)

  1. From the phone, design your button layout (fill multiple pages; leave one empty — it should hide).
  2. Drive a full session from the watch only: play, skip, sections, loop, speed, sleep timer, a voice-memo bookmark.
  3. Set the Digital Crown to scrubbing; check the deadzone (brushing it shouldn't jump position).
  4. Leave the watch off-wrist overnight; raise it next morning: right book, right position, no phantom commands?
  5. Run a Pomodoro: set 25 minutes, lower your wrist, confirm the alarm is unmissable.

Plan F — The Accessibility Pass

  1. Crank Dynamic Type to a large size: anything truncated, overlapping, unreadable?
  2. Switch the reader font to OpenDyslexic, then Lexend; adjust size and spacing.
  3. VoiceOver users: a pass over player and reader — every control labeled and operable?
  4. Enable Reduce Motion: anything still animating that shouldn't?

Plan G — The 1.0 Preview (as builds gain the new features) 🚧 Coming in 1.0

The road to 1.0 lands feature-by-feature in beta builds — each build's What to Test notes will say which of these are live. When yours has them:

  1. Insights: after a few days of normal listening, open Insights. Do the totals, streak, and per-chapter coverage match your memory of the week? Anything obviously wrong is a great report.
  2. Card Inbox: mark three passages from the transport bar (and one from the watch) without pausing; later, convert one, dismiss one, leave one. Did anything interrupt playback?
  3. Brain Dump: dictate a note from the watch mid-chapter; confirm it lands on the right book and playback never hiccuped. Promote it to a flashcard.
  4. Anki import: bring a real .apkg deck. Counts right? Scheduling sensible (mature cards not reset)? Cloze cards reported in the summary?
  5. Context Memory: opt in, bookmark something on a walk, check the place chip. Then try airplane mode — bookmarks must still save instantly. Then press Delete Location History and verify everything's gone.
  6. Study sync: with two devices on the same iCloud, create a card on one and review it on the other.

04Known limitations (current beta)

These are known — no need to report them, though opinions are welcome:

  • First auto-align is heavy. Model download (~40 MB) + Neural Engine work; phones run warm on long books. Plug in.
  • CarPlay is minimal — browse + transport only; richer templates are on the roadmap.
  • iCloud sync covers alignment anchors only so far; bookmark/flashcard/position sync ships with 1.0 (see Plan G).
  • Edition drift is real. Auto-align gets close; some books need two or three manual anchors. Expected — but tell us about books that need lots.
  • Watch review sessions can briefly show stale state if the phone app was killed mid-review; relaunching the phone app reconverges.

05Privacy during the beta

Echo's promise is unchanged in beta: no analytics, no tracking, no servers, alignment fully on-device.

One thing TestFlight itself adds: Apple's beta system shares crash reports and the feedback you choose to send with the developer, along with device/OS/build info. That's TestFlight's standard mechanism — it's how your reports reach us — not telemetry inside Echo. Full policy: privacy.html.

Thank you for testing. Every report makes the player better for the next interrupted listener. — Dan