Added initial ability to transcribe video messages.

Telegram/SourceFiles/api/api_transcribes.cpp

@@ -10,6 +10,7 @@ https://github.com/telegramdesktop/tdesktop/blob/master/LEGAL
 #include "history/history_item.h"
 #include "history/history.h"
 #include "main/main_session.h"
+#include "data/data_document.h"
 #include "data/data_session.h"
 #include "data/data_peer.h"
 #include "apiwrap.h"
@@ -30,6 +31,9 @@ void Transcribes::toggle(not_null<HistoryItem*> item) {
 		_session->data().requestItemResize(item);
 	} else if (!i->second.requestId) {
 		i->second.shown = !i->second.shown;
+		if (i->second.roundview) {
+			_session->data().requestItemViewRefresh(item);
+		}
 		_session->data().requestItemResize(item);
 	}
 }
@@ -55,6 +59,9 @@ void Transcribes::apply(const MTPDupdateTranscribedAudio &update) {
 	j->second.result = text;
 	j->second.pending = update.is_pending();
 	if (const auto item = _session->data().message(i->second)) {
+		if (j->second.roundview) {
+			_session->data().requestItemViewRefresh(item);
+		}
 		_session->data().requestItemResize(item);
 	}
 }
@@ -63,29 +70,41 @@ void Transcribes::load(not_null<HistoryItem*> item) {
 	if (!item->isHistoryEntry() || item->isLocal()) {
 		return;
 	}
+	const auto toggleRound = [](not_null<HistoryItem*> item, Entry &entry) {
+		if (const auto media = item->media()) {
+			if (const auto document = media->document()) {
+				if (document->isVideoMessage()) {
+					entry.roundview = true;
+					document->owner().requestItemViewRefresh(item);
+				}
+			}
+		}
+	};
 	const auto id = item->fullId();
 	const auto requestId = _api.request(MTPmessages_TranscribeAudio(
 		item->history()->peer->input,
 		MTP_int(item->id)
 	)).done([=](const MTPmessages_TranscribedAudio &result) {
-		result.match([&](const MTPDmessages_transcribedAudio &data) {
-			auto &entry = _map[id];
-			entry.requestId = 0;
-			entry.pending = data.is_pending();
-			entry.result = qs(data.vtext());
-			_ids.emplace(data.vtranscription_id().v, id);
-			if (const auto item = _session->data().message(id)) {
-				_session->data().requestItemResize(item);
-			}
-		});
+		const auto &data = result.data();
+		auto &entry = _map[id];
+		entry.requestId = 0;
+		entry.pending = data.is_pending();
+		entry.result = qs(data.vtext());
+		_ids.emplace(data.vtranscription_id().v, id);
+		if (const auto item = _session->data().message(id)) {
+			toggleRound(item, entry);
+			_session->data().requestItemResize(item);
+		}
 	}).fail([=](const MTP::Error &error) {
 		auto &entry = _map[id];
 		entry.requestId = 0;
 		entry.pending = false;
 		entry.failed = true;
-		if (error.type() == qstr("MSG_VOICE_TOO_LONG")) {
+		if (error.type() == u"MSG_VOICE_TOO_LONG"_q) {
 			entry.toolong = true;
-		} else if (const auto item = _session->data().message(id)) {
+		}
+		if (const auto item = _session->data().message(id)) {
+			toggleRound(item, entry);
 			_session->data().requestItemResize(item);
 		}
 	}).send();

Telegram/SourceFiles/api/api_transcribes.h

@@ -27,6 +27,7 @@ public:
 		bool failed = false;
 		bool toolong = false;
 		bool pending = false;
+		bool roundview = false;
 		mtpRequestId requestId = 0;
 	};
 

Telegram/SourceFiles/data/data_document.cpp

@@ -381,6 +381,10 @@ void DocumentData::setattributes(
 				type = data.is_round_message()
 					? RoundVideoDocument
 					: VideoDocument;
+				if (data.is_round_message()) {
+					_additional = std::make_unique<RoundData>();
+					round()->duration = data.vduration().v;
+				}
 			} else if (const auto info = sticker()) {
 				info->type = StickerType::Webm;
 			}
@@ -397,7 +401,7 @@ void DocumentData::setattributes(
 					_additional = std::make_unique<SongData>();
 				}
 			}
-			if (const auto voiceData = voice()) {
+			if (const auto voiceData = voice() ? voice() : round()) {
 				voiceData->duration = data.vduration().v;
 				voiceData->waveform = documentWaveformDecode(
 					data.vwaveform().value_or_empty());
@@ -1252,6 +1256,16 @@ const VoiceData *DocumentData::voice() const {
 	return const_cast<DocumentData*>(this)->voice();
 }
 
+RoundData *DocumentData::round() {
+	return isVideoMessage()
+		? static_cast<RoundData*>(_additional.get())
+		: nullptr;
+}
+
+const RoundData *DocumentData::round() const {
+	return const_cast<DocumentData*>(this)->round();
+}
+
 bool DocumentData::hasRemoteLocation() const {
 	return (_dc != 0 && _access != 0);
 }

Telegram/SourceFiles/data/data_document.h

@@ -90,6 +90,8 @@ struct VoiceData : public DocumentAdditionalData {
 	char wavemax = 0;
 };
 
+using RoundData = VoiceData;
+
 namespace Serialize {
 class Document;
 } // namespace Serialize;
@@ -152,6 +154,8 @@ public:
 	[[nodiscard]] const SongData *song() const;
 	[[nodiscard]] VoiceData *voice();
 	[[nodiscard]] const VoiceData *voice() const;
+	[[nodiscard]] RoundData *round();
+	[[nodiscard]] const RoundData *round() const;
 
 	void forceIsStreamedAnimation();
 	[[nodiscard]] bool isVoiceMessage() const;

Telegram/SourceFiles/data/data_media_types.cpp

@@ -1048,6 +1048,23 @@ std::unique_ptr<HistoryView::Media> MediaFile::createView(
 				_document,
 				_skipPremiumEffect,
 				replacing));
+	} else if (_document->isVideoMessage()) {
+		const auto &entry = _document->session().api().transcribes().entry(
+			parent());
+		if (!entry.requestId
+			&& entry.shown
+			&& entry.roundview
+			&& !entry.pending) {
+			return std::make_unique<HistoryView::Document>(
+				message,
+				realParent,
+				_document);
+		} else {
+			return std::make_unique<HistoryView::Gif>(
+				message,
+				realParent,
+				_document);
+		}
 	} else if (_document->isAnimation() || _document->isVideoFile()) {
 		return std::make_unique<HistoryView::Gif>(
 			message,

Telegram/SourceFiles/data/data_replies_list.cpp

@@ -839,7 +839,7 @@ std::optional<int> RepliesList::computeUnreadCountLocally(
 			end(_list),
 			wasReadTillId,
 			std::greater<>());
-		return std::max(*wasUnreadCountAfter - (till - from), 0);
+		return std::max(int(*wasUnreadCountAfter - (till - from)), 0);
 	}
 	return std::nullopt;
 }

Telegram/SourceFiles/history/view/media/history_view_document.cpp

@@ -155,7 +155,7 @@ void PaintWaveform(
 	if (const auto song = document->song()) {
 		add(FormatPlayedText(duration, duration));
 		add(FormatDurationAndSizeText(duration, document->size));
-	} else if (const auto voice = document->voice()) {
+	} else if (const auto voice = document->voice() ? document->voice() : document->round()) {
 		add(FormatPlayedText(duration, duration));
 		add(FormatDurationAndSizeText(duration, document->size));
 	} else if (document->isVideoFile()) {
@@ -174,6 +174,12 @@ Document::Document(
 	not_null<DocumentData*> document)
 : File(parent, realParent)
 , _data(document) {
+	if (_data->isVideoMessage()) {
+		const auto &entry = _data->session().api().transcribes().entry(
+			realParent);
+		_transcribedRound = entry.shown;
+	}
+
 	auto caption = createCaption();
 
 	createComponents(!caption.isEmpty());
@@ -215,7 +221,7 @@ bool Document::dataLoaded() const {
 
 void Document::createComponents(bool caption) {
 	uint64 mask = 0;
-	if (_data->isVoiceMessage()) {
+	if (_data->isVoiceMessage() || _transcribedRound) {
 		mask |= HistoryDocumentVoice::Bit();
 	} else {
 		mask |= HistoryDocumentNamed::Bit();
@@ -344,7 +350,9 @@ QSize Document::countOptimalSize() {
 	if (thumbed) {
 		accumulate_max(maxWidth, tleft + MaxStatusWidth(_data) + tright);
 	} else {
-		auto unread = _data->isVoiceMessage() ? (st::mediaUnreadSkip + st::mediaUnreadSize) : 0;
+		auto unread = (_data->isVoiceMessage() || _transcribedRound)
+			? (st::mediaUnreadSkip + st::mediaUnreadSize)
+			: 0;
 		accumulate_max(maxWidth, tleft + MaxStatusWidth(_data) + unread + _parent->skipBlockWidth() + st::msgPadding.right());
 	}
 
@@ -604,8 +612,11 @@ void Document::draw(
 	if (voice) {
 		ensureDataMediaCreated();
 
-		if (const auto voiceData = _data->voice()) {
-			if (voiceData->waveform.isEmpty()) {
+		{
+			const auto voiceData = _data->isVideoMessage()
+				? _data->round()
+				: _data->voice();
+			if (voiceData && voiceData->waveform.isEmpty()) {
 				if (loaded) {
 					Local::countVoiceWaveform(_dataMedia.get());
 				}
@@ -642,7 +653,7 @@ void Document::draw(
 
 		PaintWaveform(p,
 			context,
-			_data->voice(),
+			_transcribedRound ? _data->round() : _data->voice(),
 			namewidth + st::msgWaveformSkip,
 			progress);
 		p.restore();
@@ -1156,6 +1167,8 @@ void Document::setStatusSize(int64 newSize, TimeId realDuration) const {
 		? _data->song()->duration
 		: (_data->isVoiceMessage()
 			? _data->voice()->duration
+			: _transcribedRound
+			? _data->round()->duration
 			: -1);
 	File::setStatusSize(newSize, _data->size, duration, realDuration);
 	if (auto thumbed = Get<HistoryDocumentThumbed>()) {
@@ -1190,7 +1203,7 @@ bool Document::updateStatusText() const {
 		statusSize = Ui::FileStatusSizeReady;
 	}
 
-	if (_data->isVoiceMessage()) {
+	if (_data->isVoiceMessage() || _transcribedRound) {
 		const auto state = ::Media::Player::instance()->getState(AudioMsgId::Type::Voice);
 		if (state.id == AudioMsgId(_data, _realParent->fullId(), state.id.externalPlayId())
 			&& !::Media::Player::IsStoppedOrStopping(state.state)) {

Telegram/SourceFiles/history/view/media/history_view_document.h

@@ -174,6 +174,8 @@ private:
 
 	mutable TooltipFilename _tooltipFilename;
 
+	bool _transcribedRound = false;
+
 };
 
 bool DrawThumbnailAsSongCover(

Telegram/SourceFiles/history/view/media/history_view_gif.cpp

@@ -7,6 +7,8 @@ https://github.com/telegramdesktop/tdesktop/blob/master/LEGAL
 */
 #include "history/view/media/history_view_gif.h"
 
+#include "apiwrap.h"
+#include "api/api_transcribes.h"
 #include "lang/lang_keys.h"
 #include "mainwindow.h"
 #include "main/main_session.h"
@@ -93,6 +95,13 @@ Gif::Gif(
 			_data->loadVideoThumbnail(realParent->fullId());
 		}
 	}
+
+	_transcribe = std::make_shared<LambdaClickHandler>([=,
+			id = realParent->fullId()] {
+		if (const auto item = _data->session().data().message(id)) {
+			_data->session().api().transcribes().toggle(item);
+		}
+	});
 }
 
 Gif::~Gif() {
@@ -673,6 +682,21 @@ void Gif::draw(Painter &p, const PaintContext &context) const {
 				fastShareTop -= (st::msgDateImgDelta + st::msgDateImgPadding.y() + st::msgDateFont->height + st::msgDateImgPadding.y());
 			}
 			_parent->drawRightAction(p, context, fastShareLeft, fastShareTop, 2 * paintx + paintw);
+			if (_data->session().premium()) {
+				const auto stm = context.messageStyle();
+				PainterHighQualityEnabler hq(p);
+				p.setPen(Qt::NoPen);
+				p.setBrush(context.st->msgServiceBg());
+
+				const auto s = st::historyFastShareSize;
+				const auto r = QRect(
+					fastShareLeft,
+					fastShareTop - s - st::msgDateImgDelta,
+					s,
+					s);
+				p.drawEllipse(r);
+				context.st->historyFastTranscribeIcon().paintInCenter(p, r);
+			}
 		}
 	}
 }
@@ -976,6 +1000,17 @@ TextState Gif::textState(QPoint point, StateRequest request) const {
 			if (QRect(fastShareLeft, fastShareTop, size->width(), size->height()).contains(point)) {
 				result.link = _parent->rightActionLink();
 			}
+			if (_data->session().premium()) {
+				const auto s = st::historyFastShareSize;
+				const auto r = QRect(
+					fastShareLeft,
+					fastShareTop - s - st::msgDateImgDelta,
+					s,
+					s);
+				if (r.contains(point)) {
+					result.link = _transcribe;
+				}
+			}
 		}
 	}
 	return result;

Telegram/SourceFiles/history/view/media/history_view_gif.h

@@ -201,6 +201,8 @@ private:
 	mutable bool _thumbCacheBlurred = false;
 	mutable bool _thumbIsEllipse = false;
 
+	ClickHandlerPtr _transcribe;
+
 };
 
 } // namespace HistoryView

Telegram/SourceFiles/ui/chat/chat.style

@@ -876,6 +876,7 @@ historyFastShareBottom: 5px;
 historyFastShareIcon: icon {{ "fast_share", msgServiceFg }};
 historyGoToOriginalIcon: icon {{ "fast_to_original", msgServiceFg }};
 historyFastCommentsIcon: icon {{ "fast_comments", msgServiceFg }};
+historyFastTranscribeIcon: icon {{ "chat/voice_to_text", msgServiceFg }};
 
 historySavedFont: font(semibold 14px);
 

Telegram/SourceFiles/ui/chat/chat_style.cpp

@@ -85,6 +85,7 @@ ChatStyle::ChatStyle() {
 	make(_msgBotKbWebviewIcon, st::msgBotKbWebviewIcon);
 	make(_historyFastCommentsIcon, st::historyFastCommentsIcon);
 	make(_historyFastShareIcon, st::historyFastShareIcon);
+	make(_historyFastTranscribeIcon, st::historyFastTranscribeIcon);
 	make(_historyGoToOriginalIcon, st::historyGoToOriginalIcon);
 	make(_historyMapPoint, st::historyMapPoint);
 	make(_historyMapPointInner, st::historyMapPointInner);

Telegram/SourceFiles/ui/chat/chat_style.h

@@ -252,6 +252,9 @@ public:
 	[[nodiscard]] const style::icon &historyFastShareIcon() const {
 		return _historyFastShareIcon;
 	}
+	[[nodiscard]] const style::icon &historyFastTranscribeIcon() const {
+		return _historyFastTranscribeIcon;
+	}
 	[[nodiscard]] const style::icon &historyGoToOriginalIcon() const {
 		return _historyGoToOriginalIcon;
 	}
@@ -344,6 +347,7 @@ private:
 	style::icon _msgBotKbWebviewIcon = { Qt::Uninitialized };
 	style::icon _historyFastCommentsIcon = { Qt::Uninitialized };
 	style::icon _historyFastShareIcon = { Qt::Uninitialized };
+	style::icon _historyFastTranscribeIcon = { Qt::Uninitialized };
 	style::icon _historyGoToOriginalIcon = { Qt::Uninitialized };
 	style::icon _historyMapPoint = { Qt::Uninitialized };
 	style::icon _historyMapPointInner = { Qt::Uninitialized };