NEW: Transcribe-only mode for monolingual recordings • Settings → Translation → Passive mode → Transcribe only. Skips the translator entirely — perfect for recording a meeting, lecture, interview or podcast in a single language. You still get the polished transcript, the Apple Intelligence summary and the new Q&A on top. NEW: Ask Apple Intelligence about any saved conversation (Pro) • Open a session in History and a chat card lets you ask things like "What did they decide?", "Who said X?", "Summarise the part after the break." Multi-turn, grounded only in the transcript — no hallucinated facts. NEW: Auto-detect the source language • Pick "Auto" in the source language picker. EarPal learns what you actually speak from each Passive session and boots the next one in that language. No more tapping the picker before every conversation. NEW: Personal Voice — your voice, every language (Pro) • Translations spoken back in your own iOS Personal Voice. When the other person speaks and EarPal translates into your native language, the audio comes out in the voice you trained in Settings → Accessibility → Personal Voice. Built for people who've lost their voice to ALS or laryngectomy, and anyone who wants their iPhone to sound like them. NEW: Formatted PDF + editable RTF transcripts • Share any session as a paginated PDF (with header, footer, page breaks — ready for a journalist or lawyer handoff) or as an RTF that opens in Word, Pages, Google Docs and TextEdit for hand-corrections. Sits alongside the .vtt and .srt subtitle exports. NEW: Subtitle videos from your Gallery (Pro) • The video-subtitle flow now offers Files OR Gallery — vacation clips, screen recordings and downloaded videos drop straight in. NEW: Long-press a session in History to export • Skip opening the detail view — the full share menu is one gesture away. NEW: 9 more languages • Hebrew, Catalan, Czech, Slovak, Croatian, Hungarian, Romanian, Finnish and Malay join the roster. All ship with native iOS voices, so audio output works out of the box. EarPal now translates 30+ languages, all on-device. FIX: Older iPads now fall through to Whisper • Devices like the iPad Pro 4th gen (A12Z) where Apple doesn't ship SpeechTranscriber for English now route through Whisper automatically. No more "SpeechTranscriber does not support locale" errors. FIX: Cleaner Stop • Late translation events after pressing Stop no longer overwrite good translations with mixed-script output. FIX: Faster boot for everyday language pairs • The 1.1 GB on-device translation model only loads when your current language pair actually needs it. FIX: Personal Voice toggle now triggers the iOS permission prompt on first tap. Catch-22 unblocked. PLUS: Spanish, French, German, Italian, Japanese, Korean, Chinese, Portuguese, Arabic and Hindi see translated permission prompts, every settings screen, Glossary, Known Speakers and the full session detail.

NEW: Audio file translation (Pro) • Tap the waveform icon in the toolbar, drop in a WhatsApp voice note, meeting recording or podcast clip, and get a side-by-side transcript and translation, read aloud in any language. NEW: On-device translation packs • Settings → Languages → Translation Packs. Download Greek, Tagalog or Arabic packs (≈190 MB each) or the 100-language M2M-100 pack (≈1.1 GB) and translate even when Apple has no native model for the pair. Everything stays on your iPhone. NEW: Privacy receipts • Settings → About → Privacy receipts. Shows how many translations, listening sessions and photos you've processed, alongside the 0 bytes EarPal has sent to the cloud. Free for everyone. NEW: Playback speed (Pro) • Settings → Audio Output → Playback speed. Adjust how fast translations are read aloud in Passive mode, from 0.5× to 2.0×. NEW: Pick your speech model • Settings → Languages → Defaults → Speech pack. Choose between Standard (smaller, faster) and Highest quality (larger, more accurate for tricky languages like Greek or Thai). Fixes • Translation now works for Thai ↔ English, Turkish ↔ Dutch, Polish ↔ Arabic and other indirect pairs. Captured transcripts stay in the language you actually spoke. • Multi-speaker diarisation is more accurate in Passive mode — speakers match the transcript correctly even on single-mic recordings. • Speech-model download errors now explain what actually went wrong instead of a cryptic code. • Tapping "Manage in Settings" from a speech-pack prompt jumps straight to Languages, not the Settings root. • Mic button is ready faster on first launch — speech models load with higher priority.

NEW: Text translation sheet • Tap the speech-bubble icon in the toolbar to type or paste any phrase and translate instantly between 24 languages. • Read aloud, copy, swap-and-round-trip, auto-translate when you switch languages. • Routes through English when Apple doesn't have a direct pair (e.g. Spanish ↔ Norwegian). NEW: Speech-pack download sheet • When you pick a language EarPal needs Whisper for, a clear consent sheet appears with the one-time 150 MB download size and a live progress bar. No silent background data usage. NEW: Volume-low warning • If audio output is on but your iPhone volume is too quiet to hear translations, a banner reminds you to turn it up. Conversation mode is much smoother • Hold-to-speak engages the recording state immediately on press — no more "loading then auto-release" weirdness. • Whisper model is shared across turns: first warm-up costs once, every press after is instant. Was 30–90s recompile per press. • Late finals from Whisper aren't dropped if you press for the next turn quickly. • Single-direction fallback when the partner-side language is still loading — your side keeps working. • 60-second safety auto-stops a turn if the gesture gets interrupted. Translation reliability • Hindi, Chinese, Arabic, Norwegian, Swedish, Danish, Dutch, Polish, Russian, Turkish — all the script-tagged pairs Apple's catalog uses now correctly resolve. They were silently failing before. • Indirect pairs (Spanish ↔ Norwegian, French ↔ Polish, etc.) automatically pivot through English when both legs are installed locally. • Regional French and Portuguese variants no longer ask for Apple Intelligence — they use Apple Translation directly. • If Apple Intelligence is on and a translation pair isn't yet downloaded, EarPal translates instantly via AI while the native pack downloads in the background. • iOS 26.4 changed the install flow to require Apple Intelligence — the banner now mentions this so you know what to enable. • When Apple's translate call fails mid-translation, EarPal falls through to AI rather than showing an error. Audio handling • Silent or noisy moments no longer produce phantom translations. • Conversation mode picks the correct transcript when both languages overlap (English ↔ Japanese phonetic katakana false-positives fixed). • Better in-flight feedback: real "Translating…" indicator, accurate download progress. New settings • Keep screen on while listening passively (now ON by default) — perfect for long meetings where auto-lock used to kill the session. • Send feedback row in Settings → About — a quick form to tell us what's working and what isn't. Polish • Camera surface defaults to Photo sub-mode (Live remains a Pro feature). • "via English" badge on text-translation output when the translation pivoted through English. • Brutalist design extended throughout the new text-translation sheet. • Build cleaned up — zero compiler warnings on the latest Xcode toolchain.