diff --git a/cmd/gateway.go b/cmd/gateway.go
index e6e1c898..30d41bc0 100644
--- a/cmd/gateway.go
+++ b/cmd/gateway.go
@@ -133,7 +133,7 @@ func runGateway() {
 		tools.DetectServerIPs(context.Background())
 	}
 
-	toolsReg, execApprovalMgr, mcpMgr, sandboxMgr, browserMgr, webFetchTool, ttsTool, permPE, toolPE, dataDir, agentCfg := setupToolRegistry(cfg, workspace, providerRegistry)
+	toolsReg, execApprovalMgr, mcpMgr, sandboxMgr, browserMgr, webFetchTool, ttsTool, audioMgr, permPE, toolPE, dataDir, agentCfg := setupToolRegistry(cfg, workspace, providerRegistry)
 	if browserMgr != nil {
 		defer browserMgr.Close()
 	}
@@ -425,12 +425,12 @@ func runGateway() {
 		instanceLoader = channels.NewInstanceLoader(pgStores.ChannelInstances, pgStores.Agents, channelMgr, msgBus, pgStores.Pairing)
 		instanceLoader.SetProviderRegistry(providerRegistry)
 		instanceLoader.SetPendingCompactionConfig(cfg.Channels.PendingCompaction)
-		instanceLoader.RegisterFactory(channels.TypeTelegram, telegram.FactoryWithStores(pgStores.Agents, pgStores.ConfigPermissions, pgStores.Teams, pgStores.SubagentTasks, pgStores.PendingMessages))
-		instanceLoader.RegisterFactory(channels.TypeDiscord, discord.FactoryWithStores(pgStores.Agents, pgStores.ConfigPermissions, pgStores.PendingMessages))
-		instanceLoader.RegisterFactory(channels.TypeFeishu, feishu.FactoryWithPendingStore(pgStores.PendingMessages))
+		instanceLoader.RegisterFactory(channels.TypeTelegram, telegram.FactoryWithStoresAndAudio(pgStores.Agents, pgStores.ConfigPermissions, pgStores.Teams, pgStores.SubagentTasks, pgStores.PendingMessages, audioMgr))
+		instanceLoader.RegisterFactory(channels.TypeDiscord, discord.FactoryWithStoresAndAudio(pgStores.Agents, pgStores.ConfigPermissions, pgStores.PendingMessages, audioMgr))
+		instanceLoader.RegisterFactory(channels.TypeFeishu, feishu.FactoryWithPendingStoreAndAudio(pgStores.PendingMessages, audioMgr))
 		instanceLoader.RegisterFactory(channels.TypeZaloOA, zalo.Factory)
 		instanceLoader.RegisterFactory(channels.TypeZaloPersonal, zalopersonal.FactoryWithPendingStore(pgStores.PendingMessages))
-		instanceLoader.RegisterFactory(channels.TypeWhatsApp, whatsapp.FactoryWithDB(pgStores.DB, pgStores.PendingMessages, "pgx"))
+		instanceLoader.RegisterFactory(channels.TypeWhatsApp, whatsapp.FactoryWithDBAudio(pgStores.DB, pgStores.PendingMessages, "pgx", audioMgr, pgStores.BuiltinTools))
 		instanceLoader.RegisterFactory(channels.TypeSlack, slackchannel.FactoryWithPendingStore(pgStores.PendingMessages))
 		instanceLoader.RegisterFactory(channels.TypeFacebook, facebook.Factory)
 		instanceLoader.RegisterFactory(channels.TypePancake, pancake.Factory)
@@ -440,7 +440,7 @@ func runGateway() {
 	}
 
 	// Register config-based channels as fallback when no DB instances loaded.
-	registerConfigChannels(cfg, channelMgr, msgBus, pgStores, instanceLoader)
+	registerConfigChannels(cfg, channelMgr, msgBus, pgStores, instanceLoader, audioMgr)
 
 	// Register channels/instances/links/teams RPC methods
 	wireChannelRPCMethods(server, pgStores, channelMgr, agentRouter, msgBus, workspace)
diff --git a/cmd/gateway_channels_setup.go b/cmd/gateway_channels_setup.go
index d9d97197..df2840a3 100644
--- a/cmd/gateway_channels_setup.go
+++ b/cmd/gateway_channels_setup.go
@@ -9,6 +9,7 @@ import (
 	"github.com/google/uuid"
 
 	"github.com/nextlevelbuilder/goclaw/internal/agent"
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
 	"github.com/nextlevelbuilder/goclaw/internal/bus"
 	"github.com/nextlevelbuilder/goclaw/internal/channels"
 	"github.com/nextlevelbuilder/goclaw/internal/channels/discord"
@@ -27,7 +28,8 @@ import (
 )
 
 // registerConfigChannels registers config-based channels as fallback when no DB instances are loaded.
-func registerConfigChannels(cfg *config.Config, channelMgr *channels.Manager, msgBus *bus.MessageBus, pgStores *store.Stores, instanceLoader *channels.InstanceLoader) {
+// audioMgr is optional (nil = STT disabled for channels).
+func registerConfigChannels(cfg *config.Config, channelMgr *channels.Manager, msgBus *bus.MessageBus, pgStores *store.Stores, instanceLoader *channels.InstanceLoader, audioMgr *audio.Manager) {
 	if instanceLoader != nil {
 		return
 	}
@@ -46,7 +48,7 @@ func registerConfigChannels(cfg *config.Config, channelMgr *channels.Manager, ms
 	if cfg.Channels.Telegram.Enabled {
 		if cfg.Channels.Telegram.Token == "" {
 			recordMissingConfig(channels.TypeTelegram, "Set channels.telegram.token in config.")
-		} else if tg, err := telegram.New(cfg.Channels.Telegram, msgBus, pgStores.Pairing); err != nil {
+		} else if tg, err := telegram.New(cfg.Channels.Telegram, msgBus, pgStores.Pairing, audioMgr); err != nil {
 			channelMgr.RecordFailure(channels.TypeTelegram, "", err)
 			slog.Error("failed to initialize telegram channel", "error", err)
 		} else {
@@ -58,7 +60,7 @@ func registerConfigChannels(cfg *config.Config, channelMgr *channels.Manager, ms
 	if cfg.Channels.Discord.Enabled {
 		if cfg.Channels.Discord.Token == "" {
 			recordMissingConfig(channels.TypeDiscord, "Set channels.discord.token in config.")
-		} else if dc, err := discord.New(cfg.Channels.Discord, msgBus, nil, nil, nil, nil); err != nil {
+		} else if dc, err := discord.New(cfg.Channels.Discord, msgBus, nil, nil, nil, nil, audioMgr); err != nil {
 			channelMgr.RecordFailure(channels.TypeDiscord, "", err)
 			slog.Error("failed to initialize discord channel", "error", err)
 		} else {
@@ -72,7 +74,7 @@ func registerConfigChannels(cfg *config.Config, channelMgr *channels.Manager, ms
 		if strings.Contains(fmt.Sprintf("%T", pgStores.DB.Driver()), "sqlite") {
 			waDialect = "sqlite3"
 		}
-		wa, err := whatsapp.New(cfg.Channels.WhatsApp, msgBus, pgStores.Pairing, pgStores.DB, pgStores.PendingMessages, waDialect)
+		wa, err := whatsapp.New(cfg.Channels.WhatsApp, msgBus, pgStores.Pairing, pgStores.DB, pgStores.PendingMessages, waDialect, audioMgr, pgStores.BuiltinTools)
 		if err != nil {
 			channelMgr.RecordFailure(channels.TypeWhatsApp, "", err)
 			slog.Error("failed to initialize whatsapp channel", "error", err)
@@ -131,7 +133,7 @@ func registerConfigChannels(cfg *config.Config, channelMgr *channels.Manager, ms
 				feishu.WithAgentStore(pgStores.Agents),
 				feishu.WithConfigPermStore(pgStores.ConfigPermissions),
 			}
-			if f, err := feishu.New(cfg.Channels.Feishu, msgBus, pgStores.Pairing, nil, feishuOpts...); err != nil {
+			if f, err := feishu.New(cfg.Channels.Feishu, msgBus, pgStores.Pairing, nil, audioMgr, feishuOpts...); err != nil {
 				channelMgr.RecordFailure(channels.TypeFeishu, "", err)
 				slog.Error("failed to initialize feishu channel", "error", err)
 			} else {
diff --git a/internal/channels/discord/discord.go b/internal/channels/discord/discord.go
index 626e8f17..6d7178f2 100644
--- a/internal/channels/discord/discord.go
+++ b/internal/channels/discord/discord.go
@@ -10,6 +10,7 @@ import (
 	"github.com/bwmarrin/discordgo"
 	"github.com/google/uuid"
 
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
 	"github.com/nextlevelbuilder/goclaw/internal/bus"
 	"github.com/nextlevelbuilder/goclaw/internal/channels"
 	"github.com/nextlevelbuilder/goclaw/internal/channels/typing"
@@ -29,15 +30,17 @@ type Channel struct {
 	typingCtrls     sync.Map // channelID string → *typing.Controller
 	agentStore      store.AgentStore            // for agent key lookup (nil = writer commands disabled)
 	configPermStore store.ConfigPermissionStore // for group file writer management (nil = writer commands disabled)
+	audioMgr        *audio.Manager             // unified STT via audio.Manager (nil = no STT)
 	// pairingService, pairingDebounce, approvedGroups, groupHistory, historyLimit, requireMention
 	// are inherited from channels.BaseChannel.
 }
 
 // New creates a new Discord channel from config.
 // agentStore and configPermStore are optional (nil = writer commands disabled).
+// audioMgr is optional (nil = STT disabled).
 func New(cfg config.DiscordConfig, msgBus *bus.MessageBus, pairingSvc store.PairingStore,
 	agentStore store.AgentStore, configPermStore store.ConfigPermissionStore,
-	pendingStore store.PendingMessageStore) (*Channel, error) {
+	pendingStore store.PendingMessageStore, audioMgr *audio.Manager) (*Channel, error) {
 	session, err := discordgo.New("Bot " + cfg.Token)
 	if err != nil {
 		return nil, fmt.Errorf("create discord session: %w", err)
@@ -67,6 +70,7 @@ func New(cfg config.DiscordConfig, msgBus *bus.MessageBus, pairingSvc store.Pair
 		config:          cfg,
 		agentStore:      agentStore,
 		configPermStore: configPermStore,
+		audioMgr:        audioMgr,
 	}
 	ch.SetRequireMention(requireMention)
 	ch.SetPairingService(pairingSvc)
diff --git a/internal/channels/discord/factory.go b/internal/channels/discord/factory.go
index f2bf2a7d..9add24c6 100644
--- a/internal/channels/discord/factory.go
+++ b/internal/channels/discord/factory.go
@@ -4,6 +4,7 @@ import (
 	"encoding/json"
 	"fmt"
 
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
 	"github.com/nextlevelbuilder/goclaw/internal/bus"
 	"github.com/nextlevelbuilder/goclaw/internal/channels"
 	"github.com/nextlevelbuilder/goclaw/internal/config"
@@ -34,21 +35,26 @@ type discordInstanceConfig struct {
 // Factory creates a Discord channel from DB instance data (no extra stores).
 func Factory(name string, creds json.RawMessage, cfg json.RawMessage,
 	msgBus *bus.MessageBus, pairingSvc store.PairingStore) (channels.Channel, error) {
-	return buildChannel(name, creds, cfg, msgBus, pairingSvc, nil, nil, nil)
+	return buildChannel(name, creds, cfg, msgBus, pairingSvc, nil, nil, nil, nil)
 }
 
 // FactoryWithStores returns a ChannelFactory that includes agent, configPerm, and pending message stores.
 func FactoryWithStores(agentStore store.AgentStore, configPermStore store.ConfigPermissionStore, pendingStore store.PendingMessageStore) channels.ChannelFactory {
+	return FactoryWithStoresAndAudio(agentStore, configPermStore, pendingStore, nil)
+}
+
+// FactoryWithStoresAndAudio returns a ChannelFactory with all stores and STT support.
+func FactoryWithStoresAndAudio(agentStore store.AgentStore, configPermStore store.ConfigPermissionStore, pendingStore store.PendingMessageStore, audioMgr *audio.Manager) channels.ChannelFactory {
 	return func(name string, creds json.RawMessage, cfg json.RawMessage,
 		msgBus *bus.MessageBus, pairingSvc store.PairingStore) (channels.Channel, error) {
-		return buildChannel(name, creds, cfg, msgBus, pairingSvc, agentStore, configPermStore, pendingStore)
+		return buildChannel(name, creds, cfg, msgBus, pairingSvc, agentStore, configPermStore, pendingStore, audioMgr)
 	}
 }
 
 func buildChannel(name string, creds json.RawMessage, cfg json.RawMessage,
 	msgBus *bus.MessageBus, pairingSvc store.PairingStore,
 	agentStore store.AgentStore, configPermStore store.ConfigPermissionStore,
-	pendingStore store.PendingMessageStore) (channels.Channel, error) {
+	pendingStore store.PendingMessageStore, audioMgr *audio.Manager) (channels.Channel, error) {
 
 	var c discordCreds
 	if len(creds) > 0 {
@@ -89,7 +95,7 @@ func buildChannel(name string, creds json.RawMessage, cfg json.RawMessage,
 		dcCfg.GroupPolicy = "pairing"
 	}
 
-	ch, err := New(dcCfg, msgBus, pairingSvc, agentStore, configPermStore, pendingStore)
+	ch, err := New(dcCfg, msgBus, pairingSvc, agentStore, configPermStore, pendingStore, audioMgr)
 	if err != nil {
 		return nil, err
 	}
diff --git a/internal/channels/discord/handler.go b/internal/channels/discord/handler.go
index 9e4af222..7c466235 100644
--- a/internal/channels/discord/handler.go
+++ b/internal/channels/discord/handler.go
@@ -10,6 +10,7 @@ import (
 
 	"github.com/bwmarrin/discordgo"
 
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
 	"github.com/nextlevelbuilder/goclaw/internal/bus"
 	"github.com/nextlevelbuilder/goclaw/internal/channels"
 	"github.com/nextlevelbuilder/goclaw/internal/channels/media"
@@ -114,7 +115,18 @@ func (c *Channel) handleMessage(_ *discordgo.Session, m *discordgo.MessageCreate
 
 			switch mi.Type {
 			case media.TypeAudio, media.TypeVoice:
-				transcript, sttErr := c.transcribeAudio(context.Background(), mi.FilePath)
+				var transcript string
+				var sttErr error
+				if c.audioMgr != nil {
+					sttCtx, cancel := context.WithTimeout(ctx, 10*time.Second)
+					res, err := c.audioMgr.Transcribe(sttCtx, audio.STTInput{FilePath: mi.FilePath, MimeType: "audio/ogg"}, audio.STTOptions{})
+					cancel()
+					if err == nil && res != nil {
+						transcript = res.Text
+					} else {
+						sttErr = err
+					}
+				}
 				if sttErr != nil {
 					slog.Warn("discord: STT transcription failed",
 						"type", mi.Type, "error", sttErr,
diff --git a/internal/channels/discord/stt.go b/internal/channels/discord/stt.go
deleted file mode 100644
index ad89d826..00000000
--- a/internal/channels/discord/stt.go
+++ /dev/null
@@ -1,18 +0,0 @@
-package discord
-
-import (
-	"context"
-
-	"github.com/nextlevelbuilder/goclaw/internal/channels/media"
-)
-
-// transcribeAudio calls the shared STT proxy service with the given audio file.
-// Returns ("", nil) silently when STT is not configured or filePath is empty.
-func (c *Channel) transcribeAudio(ctx context.Context, filePath string) (string, error) {
-	return media.TranscribeAudio(ctx, media.STTConfig{
-		ProxyURL:       c.config.STTProxyURL,
-		APIKey:         c.config.STTAPIKey,
-		TenantID:       c.config.STTTenantID,
-		TimeoutSeconds: c.config.STTTimeoutSeconds,
-	}, filePath)
-}
diff --git a/internal/channels/feishu/bot.go b/internal/channels/feishu/bot.go
index 7c95eca9..931824e3 100644
--- a/internal/channels/feishu/bot.go
+++ b/internal/channels/feishu/bot.go
@@ -7,6 +7,7 @@ import (
 	"path/filepath"
 	"time"
 
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
 	"github.com/nextlevelbuilder/goclaw/internal/bus"
 	"github.com/nextlevelbuilder/goclaw/internal/channels"
 	"github.com/nextlevelbuilder/goclaw/internal/channels/media"
@@ -266,7 +267,18 @@ func (c *Channel) handleMessageEvent(ctx context.Context, event *MessageEvent) {
 
 			switch m.Type {
 			case media.TypeAudio, media.TypeVoice:
-				transcript, sttErr := c.transcribeAudio(ctx, m.FilePath)
+				var transcript string
+				var sttErr error
+				if c.audioMgr != nil {
+					sttCtx, cancel := context.WithTimeout(ctx, 10*time.Second)
+					res, err := c.audioMgr.Transcribe(sttCtx, audio.STTInput{FilePath: m.FilePath, MimeType: "audio/ogg"}, audio.STTOptions{})
+					cancel()
+					if err == nil && res != nil {
+						transcript = res.Text
+					} else {
+						sttErr = err
+					}
+				}
 				if sttErr != nil {
 					slog.Warn("feishu: STT transcription failed",
 						"type", m.Type, "error", sttErr,
diff --git a/internal/channels/feishu/factory.go b/internal/channels/feishu/factory.go
index 869a5372..1f0e9785 100644
--- a/internal/channels/feishu/factory.go
+++ b/internal/channels/feishu/factory.go
@@ -4,6 +4,7 @@ import (
 	"encoding/json"
 	"fmt"
 
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
 	"github.com/nextlevelbuilder/goclaw/internal/bus"
 	"github.com/nextlevelbuilder/goclaw/internal/channels"
 	"github.com/nextlevelbuilder/goclaw/internal/config"
@@ -100,7 +101,7 @@ func Factory(name string, creds json.RawMessage, cfg json.RawMessage,
 		fsCfg.GroupPolicy = "pairing"
 	}
 
-	ch, err := New(fsCfg, msgBus, pairingSvc, nil)
+	ch, err := New(fsCfg, msgBus, pairingSvc, nil, nil)
 	if err != nil {
 		return nil, err
 	}
@@ -111,6 +112,11 @@ func Factory(name string, creds json.RawMessage, cfg json.RawMessage,
 
 // FactoryWithPendingStore returns a ChannelFactory with persistent history support.
 func FactoryWithPendingStore(pendingStore store.PendingMessageStore) channels.ChannelFactory {
+	return FactoryWithPendingStoreAndAudio(pendingStore, nil)
+}
+
+// FactoryWithPendingStoreAndAudio returns a ChannelFactory with persistent history and STT support.
+func FactoryWithPendingStoreAndAudio(pendingStore store.PendingMessageStore, audioMgr *audio.Manager) channels.ChannelFactory {
 	return func(name string, creds json.RawMessage, cfg json.RawMessage,
 		msgBus *bus.MessageBus, pairingSvc store.PairingStore) (channels.Channel, error) {
 
@@ -165,7 +171,7 @@ func FactoryWithPendingStore(pendingStore store.PendingMessageStore) channels.Ch
 			fsCfg.GroupPolicy = "pairing"
 		}
 
-		ch, err := New(fsCfg, msgBus, pairingSvc, pendingStore)
+		ch, err := New(fsCfg, msgBus, pairingSvc, pendingStore, audioMgr)
 		if err != nil {
 			return nil, err
 		}
diff --git a/internal/channels/feishu/feishu.go b/internal/channels/feishu/feishu.go
index 2c1e167e..c91bc5e1 100644
--- a/internal/channels/feishu/feishu.go
+++ b/internal/channels/feishu/feishu.go
@@ -17,6 +17,7 @@ import (
 
 	"github.com/google/uuid"
 
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
 	"github.com/nextlevelbuilder/goclaw/internal/bus"
 	"github.com/nextlevelbuilder/goclaw/internal/channels"
 	"github.com/nextlevelbuilder/goclaw/internal/config"
@@ -49,6 +50,7 @@ type Channel struct {
 	stopCh          chan struct{}
 	httpServer      *http.Server
 	wsClient        *WSClient
+	audioMgr        *audio.Manager // unified STT via audio.Manager (nil = no STT)
 	// pairingService, pairingDebounce, approvedGroups, groupHistory, historyLimit
 	// are inherited from channels.BaseChannel.
 }
@@ -91,7 +93,8 @@ type senderCacheEntry struct {
 }
 
 // New creates a new Feishu/Lark channel.
-func New(cfg config.FeishuConfig, msgBus *bus.MessageBus, pairingSvc store.PairingStore, pendingStore store.PendingMessageStore, opts ...Option) (*Channel, error) {
+// audioMgr is optional (nil = STT disabled).
+func New(cfg config.FeishuConfig, msgBus *bus.MessageBus, pairingSvc store.PairingStore, pendingStore store.PendingMessageStore, audioMgr *audio.Manager, opts ...Option) (*Channel, error) {
 	if cfg.AppID == "" || cfg.AppSecret == "" {
 		return nil, fmt.Errorf("feishu app_id and app_secret are required")
 	}
@@ -116,6 +119,7 @@ func New(cfg config.FeishuConfig, msgBus *bus.MessageBus, pairingSvc store.Pairi
 		docCache:       newDocCache(larkDocCacheSize, larkDocCacheTTL),
 		groupAllowList: cfg.GroupAllowFrom,
 		stopCh:         make(chan struct{}),
+		audioMgr:       audioMgr,
 	}
 	ch.SetPairingService(pairingSvc)
 	ch.SetGroupHistory(channels.MakeHistory(channels.TypeFeishu, pendingStore, base.TenantID()))
diff --git a/internal/channels/feishu/stt.go b/internal/channels/feishu/stt.go
deleted file mode 100644
index 6ec5f8ba..00000000
--- a/internal/channels/feishu/stt.go
+++ /dev/null
@@ -1,18 +0,0 @@
-package feishu
-
-import (
-	"context"
-
-	"github.com/nextlevelbuilder/goclaw/internal/channels/media"
-)
-
-// transcribeAudio calls the shared STT proxy service with the given audio file.
-// Returns ("", nil) silently when STT is not configured or filePath is empty.
-func (c *Channel) transcribeAudio(ctx context.Context, filePath string) (string, error) {
-	return media.TranscribeAudio(ctx, media.STTConfig{
-		ProxyURL:       c.cfg.STTProxyURL,
-		APIKey:         c.cfg.STTAPIKey,
-		TenantID:       c.cfg.STTTenantID,
-		TimeoutSeconds: c.cfg.STTTimeoutSeconds,
-	}, filePath)
-}
diff --git a/internal/channels/telegram/channel.go b/internal/channels/telegram/channel.go
index 394ee446..392bfc27 100644
--- a/internal/channels/telegram/channel.go
+++ b/internal/channels/telegram/channel.go
@@ -14,6 +14,7 @@ import (
 	"github.com/google/uuid"
 	"github.com/mymmrac/telego"
 
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
 	"github.com/nextlevelbuilder/goclaw/internal/bus"
 	"github.com/nextlevelbuilder/goclaw/internal/channels"
 	"github.com/nextlevelbuilder/goclaw/internal/config"
@@ -43,6 +44,7 @@ type Channel struct {
 	handlerWg         sync.WaitGroup     // tracks in-flight handler goroutines for graceful shutdown
 	handlerSem        chan struct{}      // bounded semaphore for concurrent handler goroutines
 	pendingDraftID    sync.Map           // localKey string → int (draftID)
+	audioMgr          *audio.Manager    // unified STT via audio.Manager (nil = no STT)
 	// pairingService, approvedGroups, pairingDebounce, groupHistory, historyLimit, requireMention
 	// are inherited from channels.BaseChannel.
 }
@@ -85,8 +87,9 @@ func WithPendingMessageStore(s store.PendingMessageStore) Option {
 
 // New creates a new Telegram channel from config.
 // pairingSvc is optional (nil = fall back to allowlist only).
+// audioMgr is optional (nil = STT disabled).
 // Optional stores are set via Option functions.
-func New(cfg config.TelegramConfig, msgBus *bus.MessageBus, pairingSvc store.PairingStore, chanOpts ...Option) (*Channel, error) {
+func New(cfg config.TelegramConfig, msgBus *bus.MessageBus, pairingSvc store.PairingStore, audioMgr *audio.Manager, chanOpts ...Option) (*Channel, error) {
 	var botOpts []telego.BotOption
 
 	if cfg.APIServer != "" {
@@ -152,6 +155,7 @@ func New(cfg config.TelegramConfig, msgBus *bus.MessageBus, pairingSvc store.Pai
 		httpClient:  httpClient,
 		transport:   transport,
 		mentionMode: mentionMode,
+		audioMgr:    audioMgr,
 	}
 	ch.SetPairingService(pairingSvc)
 	ch.SetGroupHistory(channels.MakeHistory(channels.TypeTelegram, nil, base.TenantID()))
diff --git a/internal/channels/telegram/factory.go b/internal/channels/telegram/factory.go
index f627b4d2..9152cd36 100644
--- a/internal/channels/telegram/factory.go
+++ b/internal/channels/telegram/factory.go
@@ -4,6 +4,7 @@ import (
 	"encoding/json"
 	"fmt"
 
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
 	"github.com/nextlevelbuilder/goclaw/internal/bus"
 	"github.com/nextlevelbuilder/goclaw/internal/channels"
 	"github.com/nextlevelbuilder/goclaw/internal/config"
@@ -42,14 +43,19 @@ type telegramInstanceConfig struct {
 // Factory creates a Telegram channel from DB instance data (no extra stores).
 func Factory(name string, creds json.RawMessage, cfg json.RawMessage,
 	msgBus *bus.MessageBus, pairingSvc store.PairingStore) (channels.Channel, error) {
-	return buildChannel(name, creds, cfg, msgBus, pairingSvc)
+	return buildChannel(name, creds, cfg, msgBus, pairingSvc, nil)
 }
 
 // FactoryWithStores returns a ChannelFactory that includes optional stores via functional options.
 func FactoryWithStores(agentStore store.AgentStore, configPermStore store.ConfigPermissionStore, teamStore store.TeamStore, subagentTaskStore store.SubagentTaskStore, pendingStore store.PendingMessageStore) channels.ChannelFactory {
+	return FactoryWithStoresAndAudio(agentStore, configPermStore, teamStore, subagentTaskStore, pendingStore, nil)
+}
+
+// FactoryWithStoresAndAudio returns a ChannelFactory with all stores and STT support.
+func FactoryWithStoresAndAudio(agentStore store.AgentStore, configPermStore store.ConfigPermissionStore, teamStore store.TeamStore, subagentTaskStore store.SubagentTaskStore, pendingStore store.PendingMessageStore, audioMgr *audio.Manager) channels.ChannelFactory {
 	return func(name string, creds json.RawMessage, cfg json.RawMessage,
 		msgBus *bus.MessageBus, pairingSvc store.PairingStore) (channels.Channel, error) {
-		return buildChannel(name, creds, cfg, msgBus, pairingSvc,
+		return buildChannel(name, creds, cfg, msgBus, pairingSvc, audioMgr,
 			WithAgentStore(agentStore),
 			WithConfigPermStore(configPermStore),
 			WithTeamStore(teamStore),
@@ -60,7 +66,7 @@ func FactoryWithStores(agentStore store.AgentStore, configPermStore store.Config
 }
 
 func buildChannel(name string, creds json.RawMessage, cfg json.RawMessage,
-	msgBus *bus.MessageBus, pairingSvc store.PairingStore, opts ...Option) (channels.Channel, error) {
+	msgBus *bus.MessageBus, pairingSvc store.PairingStore, audioMgr *audio.Manager, opts ...Option) (channels.Channel, error) {
 
 	var c telegramCreds
 	if len(creds) > 0 {
@@ -117,7 +123,7 @@ func buildChannel(name string, creds json.RawMessage, cfg json.RawMessage,
 		tgCfg.GroupPolicy = "pairing"
 	}
 
-	ch, err := New(tgCfg, msgBus, pairingSvc, opts...)
+	ch, err := New(tgCfg, msgBus, pairingSvc, audioMgr, opts...)
 	if err != nil {
 		return nil, err
 	}
diff --git a/internal/channels/telegram/handlers.go b/internal/channels/telegram/handlers.go
index fcb9ac74..01586cec 100644
--- a/internal/channels/telegram/handlers.go
+++ b/internal/channels/telegram/handlers.go
@@ -10,6 +10,7 @@ import (
 	"github.com/mymmrac/telego"
 	tu "github.com/mymmrac/telego/telegoutil"
 
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
 	"github.com/nextlevelbuilder/goclaw/internal/bus"
 	"github.com/nextlevelbuilder/goclaw/internal/channels"
 	"github.com/nextlevelbuilder/goclaw/internal/channels/typing"
@@ -407,7 +408,18 @@ func (c *Channel) handleMessage(ctx context.Context, update telego.Update) {
 			m := &mediaList[i]
 			switch m.Type {
 			case "audio", "voice":
-				transcript, sttErr := c.transcribeAudio(ctx, m.FilePath)
+				var transcript string
+				var sttErr error
+				if c.audioMgr != nil {
+					sttCtx, cancel := context.WithTimeout(ctx, 10*time.Second)
+					res, err := c.audioMgr.Transcribe(sttCtx, audio.STTInput{FilePath: m.FilePath, MimeType: "audio/ogg"}, audio.STTOptions{})
+					cancel()
+					if err == nil && res != nil {
+						transcript = res.Text
+					} else {
+						sttErr = err
+					}
+				}
 				if sttErr != nil {
 					slog.Warn("telegram: STT transcription failed",
 						"type", m.Type, "error", sttErr)
diff --git a/internal/channels/telegram/stt.go b/internal/channels/telegram/stt.go
deleted file mode 100644
index df9d005d..00000000
--- a/internal/channels/telegram/stt.go
+++ /dev/null
@@ -1,19 +0,0 @@
-package telegram
-
-import (
-	"context"
-
-	"github.com/nextlevelbuilder/goclaw/internal/channels/media"
-)
-
-// transcribeAudio calls the configured STT proxy service with the given audio file and returns
-// the transcribed text. Returns ("", nil) silently when STT is not configured or filePath is empty.
-// Delegates to the shared media.TranscribeAudio implementation.
-func (c *Channel) transcribeAudio(ctx context.Context, filePath string) (string, error) {
-	return media.TranscribeAudio(ctx, media.STTConfig{
-		ProxyURL:       c.config.STTProxyURL,
-		APIKey:         c.config.STTAPIKey,
-		TenantID:       c.config.STTTenantID,
-		TimeoutSeconds: c.config.STTTimeoutSeconds,
-	}, filePath)
-}
diff --git a/internal/channels/telegram/stt_test.go b/internal/channels/telegram/stt_test.go
deleted file mode 100644
index 236163a0..00000000
--- a/internal/channels/telegram/stt_test.go
+++ /dev/null
@@ -1,312 +0,0 @@
-package telegram
-
-import (
-	"context"
-	"encoding/json"
-	"net/http"
-	"net/http/httptest"
-	"os"
-	"strings"
-	"testing"
-
-	"github.com/nextlevelbuilder/goclaw/internal/config"
-)
-
-// sttTestResponse mirrors the STT proxy JSON response for test assertions.
-type sttTestResponse struct {
-	Transcript string `json:"transcript"`
-}
-
-// sttEndpoint is the expected STT endpoint path.
-const sttEndpoint = "/transcribe_audio"
-
-// newChannelWithSTT is a minimal Channel stub for STT unit tests.
-// It skips bot initialisation (which requires a real Telegram token).
-func newChannelWithSTT(cfg config.TelegramConfig) *Channel {
-	return &Channel{config: cfg}
-}
-
-// writeTempAudio writes a fake audio file and returns its path.
-// The caller is responsible for removing the file after the test.
-func writeTempAudio(t *testing.T, content string) string {
-	t.Helper()
-	f, err := os.CreateTemp("", "stt_test_*.ogg")
-	if err != nil {
-		t.Fatalf("create temp audio file: %v", err)
-	}
-	if _, err := f.WriteString(content); err != nil {
-		t.Fatalf("write temp audio file: %v", err)
-	}
-	f.Close()
-	return f.Name()
-}
-
-// --- transcribeAudio unit tests ---
-
-// TestTranscribeAudio_NoProxy verifies that when STTProxyURL is empty, the
-// function returns ("", nil) without making any HTTP call.
-func TestTranscribeAudio_NoProxy(t *testing.T) {
-	c := newChannelWithSTT(config.TelegramConfig{})
-	transcript, err := c.transcribeAudio(context.Background(), "/any/file.ogg")
-	if err != nil {
-		t.Fatalf("expected nil error, got: %v", err)
-	}
-	if transcript != "" {
-		t.Fatalf("expected empty transcript, got: %q", transcript)
-	}
-}
-
-// TestTranscribeAudio_EmptyFilePath verifies that an empty filePath is a silent
-// no-op even when STT is configured.
-func TestTranscribeAudio_EmptyFilePath(t *testing.T) {
-	c := newChannelWithSTT(config.TelegramConfig{
-		STTProxyURL: "https://stt.example.com",
-	})
-	transcript, err := c.transcribeAudio(context.Background(), "")
-	if err != nil {
-		t.Fatalf("expected nil error, got: %v", err)
-	}
-	if transcript != "" {
-		t.Fatalf("expected empty transcript, got: %q", transcript)
-	}
-}
-
-// TestTranscribeAudio_MissingFile verifies that a non-existent file path returns
-// an error (not a silent empty result).
-func TestTranscribeAudio_MissingFile(t *testing.T) {
-	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		// Should never be reached.
-		t.Error("unexpected HTTP call for missing file")
-	}))
-	defer srv.Close()
-
-	c := newChannelWithSTT(config.TelegramConfig{STTProxyURL: srv.URL})
-	_, err := c.transcribeAudio(context.Background(), "/nonexistent/file.ogg")
-	if err == nil {
-		t.Fatal("expected an error for missing file, got nil")
-	}
-}
-
-// TestTranscribeAudio_Success verifies the happy path: a real HTTP server returns
-// {"transcript": "hello world"} and the function returns that string.
-func TestTranscribeAudio_Success(t *testing.T) {
-	audioFile := writeTempAudio(t, "fake-ogg-bytes")
-	defer os.Remove(audioFile)
-
-	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		// Verify endpoint path.
-		if r.URL.Path != sttEndpoint {
-			t.Errorf("unexpected path: %s", r.URL.Path)
-		}
-		// Verify method.
-		if r.Method != http.MethodPost {
-			t.Errorf("expected POST, got %s", r.Method)
-		}
-		// Verify multipart body contains a "file" field.
-		if err := r.ParseMultipartForm(1 << 20); err != nil {
-			t.Errorf("parse multipart: %v", err)
-		}
-		if _, _, err := r.FormFile("file"); err != nil {
-			t.Errorf("expected 'file' field in multipart form: %v", err)
-		}
-
-		w.Header().Set("Content-Type", "application/json")
-		json.NewEncoder(w).Encode(sttTestResponse{Transcript: "hello world"})
-	}))
-	defer srv.Close()
-
-	c := newChannelWithSTT(config.TelegramConfig{STTProxyURL: srv.URL})
-	transcript, err := c.transcribeAudio(context.Background(), audioFile)
-	if err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-	if transcript != "hello world" {
-		t.Errorf("expected %q, got %q", "hello world", transcript)
-	}
-}
-
-// TestTranscribeAudio_BearerToken verifies that STTAPIKey is sent as an
-// Authorization: Bearer header.
-func TestTranscribeAudio_BearerToken(t *testing.T) {
-	audioFile := writeTempAudio(t, "fake-ogg-bytes")
-	defer os.Remove(audioFile)
-
-	const wantKey = "super-secret-key"
-	var gotAuth string
-
-	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		gotAuth = r.Header.Get("Authorization")
-		w.Header().Set("Content-Type", "application/json")
-		json.NewEncoder(w).Encode(sttTestResponse{Transcript: "ok"})
-	}))
-	defer srv.Close()
-
-	c := newChannelWithSTT(config.TelegramConfig{
-		STTProxyURL: srv.URL,
-		STTAPIKey:   wantKey,
-	})
-	if _, err := c.transcribeAudio(context.Background(), audioFile); err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-	if gotAuth != "Bearer "+wantKey {
-		t.Errorf("expected Authorization %q, got %q", "Bearer "+wantKey, gotAuth)
-	}
-}
-
-// TestTranscribeAudio_NoAuthHeader verifies that no Authorization header is sent
-// when STTAPIKey is empty.
-func TestTranscribeAudio_NoAuthHeader(t *testing.T) {
-	audioFile := writeTempAudio(t, "fake-ogg-bytes")
-	defer os.Remove(audioFile)
-
-	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		if auth := r.Header.Get("Authorization"); auth != "" {
-			t.Errorf("expected no Authorization header, got %q", auth)
-		}
-		w.Header().Set("Content-Type", "application/json")
-		json.NewEncoder(w).Encode(sttTestResponse{Transcript: "ok"})
-	}))
-	defer srv.Close()
-
-	c := newChannelWithSTT(config.TelegramConfig{STTProxyURL: srv.URL})
-	if _, err := c.transcribeAudio(context.Background(), audioFile); err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-}
-
-// TestTranscribeAudio_TenantID verifies that STTTenantID is forwarded as a
-// multipart "tenant_id" field when set.
-func TestTranscribeAudio_TenantID(t *testing.T) {
-	audioFile := writeTempAudio(t, "fake-ogg-bytes")
-	defer os.Remove(audioFile)
-
-	const wantTenant = "acme-corp"
-	var gotTenant string
-
-	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		if err := r.ParseMultipartForm(1 << 20); err == nil {
-			gotTenant = r.FormValue("tenant_id")
-		}
-		w.Header().Set("Content-Type", "application/json")
-		json.NewEncoder(w).Encode(sttTestResponse{Transcript: "ok"})
-	}))
-	defer srv.Close()
-
-	c := newChannelWithSTT(config.TelegramConfig{
-		STTProxyURL: srv.URL,
-		STTTenantID: wantTenant,
-	})
-	if _, err := c.transcribeAudio(context.Background(), audioFile); err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-	if gotTenant != wantTenant {
-		t.Errorf("expected tenant_id %q, got %q", wantTenant, gotTenant)
-	}
-}
-
-// TestTranscribeAudio_NoTenantField verifies that when STTTenantID is empty, the
-// multipart form does NOT include a "tenant_id" field.
-func TestTranscribeAudio_NoTenantField(t *testing.T) {
-	audioFile := writeTempAudio(t, "fake-ogg-bytes")
-	defer os.Remove(audioFile)
-
-	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		if err := r.ParseMultipartForm(1 << 20); err == nil {
-			if tid := r.FormValue("tenant_id"); tid != "" {
-				t.Errorf("expected no tenant_id field, got %q", tid)
-			}
-		}
-		w.Header().Set("Content-Type", "application/json")
-		json.NewEncoder(w).Encode(sttTestResponse{Transcript: "ok"})
-	}))
-	defer srv.Close()
-
-	c := newChannelWithSTT(config.TelegramConfig{STTProxyURL: srv.URL})
-	if _, err := c.transcribeAudio(context.Background(), audioFile); err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-}
-
-// TestTranscribeAudio_UpstreamError verifies that a non-200 response is surfaced
-// as an error (not silently swallowed).
-func TestTranscribeAudio_UpstreamError(t *testing.T) {
-	audioFile := writeTempAudio(t, "fake-ogg-bytes")
-	defer os.Remove(audioFile)
-
-	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		http.Error(w, "service unavailable", http.StatusServiceUnavailable)
-	}))
-	defer srv.Close()
-
-	c := newChannelWithSTT(config.TelegramConfig{STTProxyURL: srv.URL})
-	_, err := c.transcribeAudio(context.Background(), audioFile)
-	if err == nil {
-		t.Fatal("expected error for non-200 response, got nil")
-	}
-	if !strings.Contains(err.Error(), "503") {
-		t.Errorf("expected error to mention status 503, got: %v", err)
-	}
-}
-
-// TestTranscribeAudio_InvalidJSON verifies that a 200 response with malformed
-// JSON is returned as an error.
-func TestTranscribeAudio_InvalidJSON(t *testing.T) {
-	audioFile := writeTempAudio(t, "fake-ogg-bytes")
-	defer os.Remove(audioFile)
-
-	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		w.Header().Set("Content-Type", "application/json")
-		w.Write([]byte(`not-json`))
-	}))
-	defer srv.Close()
-
-	c := newChannelWithSTT(config.TelegramConfig{STTProxyURL: srv.URL})
-	_, err := c.transcribeAudio(context.Background(), audioFile)
-	if err == nil {
-		t.Fatal("expected error for invalid JSON, got nil")
-	}
-}
-
-// TestTranscribeAudio_EmptyTranscript verifies that a 200 response with an empty
-// transcript field returns ("", nil) — not an error.
-func TestTranscribeAudio_EmptyTranscript(t *testing.T) {
-	audioFile := writeTempAudio(t, "fake-ogg-bytes")
-	defer os.Remove(audioFile)
-
-	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		w.Header().Set("Content-Type", "application/json")
-		json.NewEncoder(w).Encode(sttTestResponse{Transcript: ""})
-	}))
-	defer srv.Close()
-
-	c := newChannelWithSTT(config.TelegramConfig{STTProxyURL: srv.URL})
-	transcript, err := c.transcribeAudio(context.Background(), audioFile)
-	if err != nil {
-		t.Fatalf("unexpected error: %v", err)
-	}
-	if transcript != "" {
-		t.Errorf("expected empty transcript, got %q", transcript)
-	}
-}
-
-// TestTranscribeAudio_ContextCancelled verifies that a cancelled context causes
-// the HTTP call to fail fast.
-func TestTranscribeAudio_ContextCancelled(t *testing.T) {
-	audioFile := writeTempAudio(t, "fake-ogg-bytes")
-	defer os.Remove(audioFile)
-
-	// Server that blocks until the test is done — ensures the cancel fires first.
-	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		<-r.Context().Done()
-	}))
-	defer srv.Close()
-
-	ctx, cancel := context.WithCancel(context.Background())
-	cancel() // cancel immediately
-
-	c := newChannelWithSTT(config.TelegramConfig{STTProxyURL: srv.URL})
-	_, err := c.transcribeAudio(ctx, audioFile)
-	if err == nil {
-		t.Fatal("expected error for cancelled context, got nil")
-	}
-}
diff --git a/internal/channels/whatsapp/factory.go b/internal/channels/whatsapp/factory.go
index e49801eb..78316aa4 100644
--- a/internal/channels/whatsapp/factory.go
+++ b/internal/channels/whatsapp/factory.go
@@ -5,6 +5,7 @@ import (
 	"encoding/json"
 	"fmt"
 
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
 	"github.com/nextlevelbuilder/goclaw/internal/bus"
 	"github.com/nextlevelbuilder/goclaw/internal/channels"
 	"github.com/nextlevelbuilder/goclaw/internal/config"
@@ -24,6 +25,13 @@ type whatsappInstanceConfig struct {
 // FactoryWithDB returns a ChannelFactory with DB access for whatsmeow auth state.
 // dialect must be "pgx" (PostgreSQL) or "sqlite3" (SQLite/desktop).
 func FactoryWithDB(db *sql.DB, pendingStore store.PendingMessageStore, dialect string) channels.ChannelFactory {
+	return FactoryWithDBAudio(db, pendingStore, dialect, nil, nil)
+}
+
+// FactoryWithDBAudio returns a ChannelFactory with DB access, STT support, and builtin-tools store
+// for reading stt.whatsapp_enabled opt-in setting per message.
+func FactoryWithDBAudio(db *sql.DB, pendingStore store.PendingMessageStore, dialect string,
+	audioMgr *audio.Manager, builtinToolStore store.BuiltinToolStore) channels.ChannelFactory {
 	return func(name string, creds json.RawMessage, cfg json.RawMessage,
 		msgBus *bus.MessageBus, pairingSvc store.PairingStore) (channels.Channel, error) {
 
@@ -64,7 +72,7 @@ func FactoryWithDB(db *sql.DB, pendingStore store.PendingMessageStore, dialect s
 			waCfg.GroupPolicy = "pairing"
 		}
 
-		ch, err := New(waCfg, msgBus, pairingSvc, db, pendingStore, dialect)
+		ch, err := New(waCfg, msgBus, pairingSvc, db, pendingStore, dialect, audioMgr, builtinToolStore)
 		if err != nil {
 			return nil, err
 		}
diff --git a/internal/channels/whatsapp/inbound.go b/internal/channels/whatsapp/inbound.go
index 0deb3f0a..9414ef00 100644
--- a/internal/channels/whatsapp/inbound.go
+++ b/internal/channels/whatsapp/inbound.go
@@ -111,6 +111,21 @@ func (c *Channel) handleIncomingMessage(evt *events.Message) {
 		metadata["user_name"] = evt.Info.PushName
 	}
 
+	// STT: transcribe audio items (opt-in via builtin_tools[stt].settings.whatsapp_enabled,
+	// default false per Decision 6 — enabling breaks E2E encryption for voice messages).
+	waSttSettings := c.loadSTTSettings(ctx)
+	locale := "" // i18n.T falls back to English when locale is empty
+	for i := range mediaList {
+		m := &mediaList[i]
+		if m.Type == media.TypeAudio || m.Type == media.TypeVoice {
+			mimeType := m.ContentType
+			if mimeType == "" {
+				mimeType = "audio/ogg"
+			}
+			m.Transcript = c.transcribeVoice(ctx, m.FilePath, mimeType, locale, waSttSettings)
+		}
+	}
+
 	// Build media tags and bus.MediaFile list.
 	var mediaFiles []bus.MediaFile
 	if len(mediaList) > 0 {
diff --git a/internal/channels/whatsapp/stt.go b/internal/channels/whatsapp/stt.go
new file mode 100644
index 00000000..4c90495f
--- /dev/null
+++ b/internal/channels/whatsapp/stt.go
@@ -0,0 +1,65 @@
+package whatsapp
+
+import (
+	"context"
+	"encoding/json"
+	"log/slog"
+	"time"
+
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
+	"github.com/nextlevelbuilder/goclaw/internal/i18n"
+)
+
+// sttSettings is the minimal contract Phase 5 needs from builtin_tools[stt].settings.
+// whatsapp_enabled defaults to false (opt-in) per Decision 6 — admin must explicitly
+// enable it, acknowledging that WhatsApp STT breaks E2E encryption for voice messages.
+type sttSettings struct {
+	WhatsappEnabled bool `json:"whatsapp_enabled"`
+}
+
+// loadSTTSettings fetches stt builtin-tool settings and parses whatsapp_enabled.
+// Returns opt-out (false) on any failure: nil store, DB error, parse error, or missing key.
+// Called per voice message — rare enough that per-message DB read is acceptable.
+func (c *Channel) loadSTTSettings(ctx context.Context) sttSettings {
+	if c.builtinToolStore == nil {
+		return sttSettings{WhatsappEnabled: false}
+	}
+	raw, err := c.builtinToolStore.GetSettings(ctx, "stt")
+	if err != nil {
+		slog.Debug("whatsapp: stt settings fetch failed, defaulting to opt-out", "error", err)
+		return sttSettings{WhatsappEnabled: false}
+	}
+	if len(raw) == 0 {
+		return sttSettings{WhatsappEnabled: false}
+	}
+	var s sttSettings
+	if err := json.Unmarshal(raw, &s); err != nil {
+		slog.Debug("whatsapp: stt settings parse failed, defaulting to opt-out", "error", err)
+		return sttSettings{WhatsappEnabled: false}
+	}
+	return s
+}
+
+// transcribeVoice converts a downloaded voice file to text via audio.Manager.
+// Returns i18n fallback "[Voice message]" when:
+//   - settings.WhatsappEnabled == false (default; opt-in per Decision 6)
+//   - audioMgr is nil
+//   - Transcribe fails / times out (10s wall clock)
+func (c *Channel) transcribeVoice(ctx context.Context, filePath, mimeType, locale string, settings sttSettings) string {
+	fallback := i18n.T(locale, i18n.MsgVoiceMessageFallback)
+	if !settings.WhatsappEnabled {
+		return fallback
+	}
+	if c.audioMgr == nil {
+		return fallback
+	}
+	sttCtx, cancel := context.WithTimeout(ctx, 10*time.Second)
+	defer cancel()
+	res, err := c.audioMgr.Transcribe(sttCtx, audio.STTInput{FilePath: filePath, MimeType: mimeType}, audio.STTOptions{})
+	if err != nil || res == nil {
+		slog.Warn("whatsapp: stt failed or timed out", "error", err, "file", filePath)
+		return fallback
+	}
+	slog.Info("audio.stt.whatsapp_invoked", "duration_secs", res.Duration, "lang", res.Language)
+	return res.Text
+}
diff --git a/internal/channels/whatsapp/stt_test.go b/internal/channels/whatsapp/stt_test.go
new file mode 100644
index 00000000..c1326099
--- /dev/null
+++ b/internal/channels/whatsapp/stt_test.go
@@ -0,0 +1,163 @@
+package whatsapp
+
+import (
+	"context"
+	"errors"
+	"testing"
+	"time"
+
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
+	"github.com/nextlevelbuilder/goclaw/internal/i18n"
+)
+
+// stubSTTProvider is a minimal in-process STTProvider for unit tests.
+type stubSTTProvider struct {
+	name    string
+	result  *audio.TranscriptResult
+	err     error
+	delay   time.Duration // simulate slow provider
+	called  int
+}
+
+func (s *stubSTTProvider) Name() string { return s.name }
+
+func (s *stubSTTProvider) Transcribe(ctx context.Context, in audio.STTInput, opts audio.STTOptions) (*audio.TranscriptResult, error) {
+	s.called++
+	if s.delay > 0 {
+		select {
+		case <-time.After(s.delay):
+		case <-ctx.Done():
+			return nil, ctx.Err()
+		}
+	}
+	return s.result, s.err
+}
+
+// newTestChannel builds a minimal Channel with audioMgr for unit tests.
+func newTestChannelWithAudio(audioMgr *audio.Manager) *Channel {
+	return &Channel{audioMgr: audioMgr}
+}
+
+// newAudioMgr creates an audio.Manager with one stub STT provider.
+func newAudioMgr(stub *stubSTTProvider) *audio.Manager {
+	mgr := audio.NewManager(audio.ManagerConfig{})
+	mgr.RegisterSTT(stub)
+	mgr.SetSTTChain([]string{stub.Name()})
+	return mgr
+}
+
+// Case 1: opt-in enabled, success — transcribeVoice calls Transcribe and returns transcript.
+func TestTranscribeVoice_OptInSuccess_CallsManager(t *testing.T) {
+	stub := &stubSTTProvider{name: "stub", result: &audio.TranscriptResult{Text: "hello world"}}
+	mgr := newAudioMgr(stub)
+	ch := newTestChannelWithAudio(mgr)
+	settings := sttSettings{WhatsappEnabled: true}
+
+	got := ch.transcribeVoice(context.Background(), "/tmp/voice.ogg", "audio/ogg", "en", settings)
+
+	if stub.called != 1 {
+		t.Errorf("expected Transcribe called once, got %d", stub.called)
+	}
+	if got != "hello world" {
+		t.Errorf("expected %q, got %q", "hello world", got)
+	}
+}
+
+// Case 2: opt-in + Transcribe succeeds → returns transcript text.
+func TestTranscribeVoice_Success_ReturnsTranscript(t *testing.T) {
+	stub := &stubSTTProvider{name: "stub", result: &audio.TranscriptResult{Text: "test transcript"}}
+	mgr := newAudioMgr(stub)
+	ch := newTestChannelWithAudio(mgr)
+	settings := sttSettings{WhatsappEnabled: true}
+
+	got := ch.transcribeVoice(context.Background(), "/tmp/audio.ogg", "audio/ogg", "", settings)
+
+	if got != "test transcript" {
+		t.Errorf("expected %q, got %q", "test transcript", got)
+	}
+}
+
+// Case 3: STT error → fallback "[Voice message]" returned.
+func TestTranscribeVoice_STTError_ReturnsFallback(t *testing.T) {
+	stub := &stubSTTProvider{name: "stub", err: errors.New("provider unavailable")}
+	mgr := newAudioMgr(stub)
+	ch := newTestChannelWithAudio(mgr)
+	settings := sttSettings{WhatsappEnabled: true}
+
+	got := ch.transcribeVoice(context.Background(), "/tmp/voice.ogg", "audio/ogg", "en", settings)
+
+	fallback := i18n.T("en", i18n.MsgVoiceMessageFallback)
+	if got != fallback {
+		t.Errorf("expected fallback %q, got %q", fallback, got)
+	}
+}
+
+// Case 4: ctx DeadlineExceeded → fallback, no crash.
+func TestTranscribeVoice_ContextDeadlineExceeded_ReturnsFallback(t *testing.T) {
+	// Stub that blocks until ctx done — simulates slow Scribe.
+	stub := &stubSTTProvider{name: "stub", delay: 5 * time.Second}
+	mgr := newAudioMgr(stub)
+	ch := newTestChannelWithAudio(mgr)
+	settings := sttSettings{WhatsappEnabled: true}
+
+	ctx, cancel := context.WithTimeout(context.Background(), 50*time.Millisecond)
+	defer cancel()
+
+	got := ch.transcribeVoice(ctx, "/tmp/voice.ogg", "audio/ogg", "en", settings)
+
+	fallback := i18n.T("en", i18n.MsgVoiceMessageFallback)
+	if got != fallback {
+		t.Errorf("expected fallback %q, got %q", fallback, got)
+	}
+}
+
+// Case 5: Manager has zero STT providers → ErrAllSTTProvidersFailed → fallback.
+func TestTranscribeVoice_NoProviders_ReturnsFallback(t *testing.T) {
+	mgr := audio.NewManager(audio.ManagerConfig{})
+	// No providers registered — Transcribe returns ErrAllSTTProvidersFailed.
+	ch := newTestChannelWithAudio(mgr)
+	settings := sttSettings{WhatsappEnabled: true}
+
+	got := ch.transcribeVoice(context.Background(), "/tmp/voice.ogg", "audio/ogg", "", settings)
+
+	fallback := i18n.T("", i18n.MsgVoiceMessageFallback)
+	if got != fallback {
+		t.Errorf("expected fallback %q, got %q", fallback, got)
+	}
+}
+
+// Case 6: ctx Done before 10s wall clock → Transcribe aborts, returns fallback.
+func TestTranscribeVoice_CtxDoneBeforeTimeout_AbortsCleanly(t *testing.T) {
+	stub := &stubSTTProvider{name: "stub", delay: 30 * time.Second}
+	mgr := newAudioMgr(stub)
+	ch := newTestChannelWithAudio(mgr)
+	settings := sttSettings{WhatsappEnabled: true}
+
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel() // immediately cancelled
+
+	got := ch.transcribeVoice(ctx, "/tmp/voice.ogg", "audio/ogg", "vi", settings)
+
+	fallback := i18n.T("vi", i18n.MsgVoiceMessageFallback)
+	if got != fallback {
+		t.Errorf("expected fallback %q, got %q", fallback, got)
+	}
+}
+
+// Case 7: whatsapp_enabled=false (default opt-out) → Transcribe never called, fallback only.
+func TestTranscribeVoice_OptOut_NoTranscribeCall_ReturnsFallback(t *testing.T) {
+	stub := &stubSTTProvider{name: "stub", result: &audio.TranscriptResult{Text: "should not appear"}}
+	mgr := newAudioMgr(stub)
+	ch := newTestChannelWithAudio(mgr)
+	settings := sttSettings{WhatsappEnabled: false} // opt-out (default)
+
+	got := ch.transcribeVoice(context.Background(), "/tmp/voice.ogg", "audio/ogg", "en", settings)
+
+	if stub.called != 0 {
+		t.Errorf("expected Transcribe NOT called, got %d calls", stub.called)
+	}
+	fallback := i18n.T("en", i18n.MsgVoiceMessageFallback)
+	if got != fallback {
+		t.Errorf("expected fallback %q, got %q", fallback, got)
+	}
+}
diff --git a/internal/channels/whatsapp/whatsapp.go b/internal/channels/whatsapp/whatsapp.go
index 2964cdb7..cbd8afd9 100644
--- a/internal/channels/whatsapp/whatsapp.go
+++ b/internal/channels/whatsapp/whatsapp.go
@@ -13,8 +13,8 @@ import (
 	"go.mau.fi/whatsmeow/store/sqlstore"
 	"go.mau.fi/whatsmeow/types"
 	"go.mau.fi/whatsmeow/types/events"
-	"google.golang.org/protobuf/proto"
 
+	"github.com/nextlevelbuilder/goclaw/internal/audio"
 	"github.com/nextlevelbuilder/goclaw/internal/bus"
 	"github.com/nextlevelbuilder/goclaw/internal/channels"
 	"github.com/nextlevelbuilder/goclaw/internal/config"
@@ -28,7 +28,7 @@ const (
 
 func init() {
 	// Set device name shown in WhatsApp's "Linked Devices" screen (once at package init).
-	wastore.DeviceProps.Os = proto.String("GoClaw")
+	wastore.DeviceProps.Os = new("GoClaw")
 }
 
 // Channel connects directly to WhatsApp via go.mau.fi/whatsmeow.
@@ -41,7 +41,9 @@ type Channel struct {
 	mu        sync.Mutex
 	ctx       context.Context
 	cancel    context.CancelFunc
-	parentCtx context.Context // stored from Start() for Reauth() context chain
+	parentCtx        context.Context       // stored from Start() for Reauth() context chain
+	audioMgr         *audio.Manager        // unified STT via audio.Manager (nil = no STT)
+	builtinToolStore store.BuiltinToolStore // reads stt settings (whatsapp_enabled) per voice message; nil = opt-out
 
 	// QR state
 	lastQRMu        sync.RWMutex
@@ -58,7 +60,6 @@ type Channel struct {
 	// pairingService, pairingDebounce, approvedGroups, groupHistory are inherited from channels.BaseChannel.
 }
 
-
 // GetLastQRB64 returns the most recent QR PNG (base64).
 func (c *Channel) GetLastQRB64() string {
 	c.lastQRMu.RLock()
@@ -82,9 +83,12 @@ func (c *Channel) cacheQR(pngB64 string) {
 
 // New creates a new WhatsApp channel backed by whatsmeow.
 // dialect must be "pgx" (PostgreSQL) or "sqlite3" (SQLite/desktop).
+// audioMgr is optional (nil = STT disabled).
+// builtinToolStore is optional (nil = STT permanently opt-out regardless of admin toggle).
 func New(cfg config.WhatsAppConfig, msgBus *bus.MessageBus,
 	pairingSvc store.PairingStore, db *sql.DB,
-	pendingStore store.PendingMessageStore, dialect string) (*Channel, error) {
+	pendingStore store.PendingMessageStore, dialect string, audioMgr *audio.Manager,
+	builtinToolStore store.BuiltinToolStore) (*Channel, error) {
 
 	base := channels.NewBaseChannel(channels.TypeWhatsApp, msgBus, cfg.AllowFrom)
 	base.ValidatePolicy(cfg.DMPolicy, cfg.GroupPolicy)
@@ -95,9 +99,11 @@ func New(cfg config.WhatsAppConfig, msgBus *bus.MessageBus,
 	}
 
 	ch := &Channel{
-		BaseChannel: base,
-		config:      cfg,
-		container:   container,
+		BaseChannel:      base,
+		config:           cfg,
+		container:        container,
+		audioMgr:         audioMgr,
+		builtinToolStore: builtinToolStore,
 	}
 	ch.SetPairingService(pairingSvc)
 	ch.SetGroupHistory(channels.MakeHistory("whatsapp", pendingStore, base.TenantID()))
@@ -117,7 +123,6 @@ func (c *Channel) Start(ctx context.Context) error {
 		return fmt.Errorf("whatsapp get device: %w", err)
 	}
 
-
 	c.client = whatsmeow.NewClient(deviceStore, nil)
 	c.client.AddEventHandler(c.handleEvent)