TalkToMeBaby

TalkToMeBaby is a Speechify-inspired read-aloud extension for VS Code. It reads your prose files aloud with a dedicated Reader panel, word-level karaoke highlighting, and click-to-jump navigation. On macOS it defaults to the built-in say command (official, fully offline, no data leaves the machine); on Windows and Linux it defaults to the free unofficial Edge TTS provider. Any provider can be chosen explicitly. Supported providers: Edge TTS, ElevenLabs, macOS say, and Sarvam AI.

TalkToMeBaby is independent and is not affiliated with, endorsed by, or sponsored by Speechify.

Use of TalkToMeBaby is at your own risk. You are responsible for choosing a provider, complying with provider terms, ensuring you have rights to send or synthesize the text you read, and protecting confidential or regulated content. See DISCLAIMER.md and PRIVACY.md.

Core features:

• Reader panel - rendered reading view with dual-layer highlighting: sentence band + moving word sweep
• Click any word to jump playback there instantly; alt+j jumps from the source editor to the cursor
• Speed 0.5-2x with pitch preserved, no re-synthesis; presets + fine slider, persisted across sessions
• Disk cache (default 200 MB) so re-reads cost no API credits

Supported file types: .md, .mdx, .txt, .rst, .org, .tex, .adoc

Install

From the VS Code Marketplace

Install from the VS Code Marketplace, or from VS Code:

1. Open the Extensions view
2. Search for TalkToMeBaby
3. Click Install

You can also install it from the command line:

code --install-extension rishmadaan.talktomebaby

From VSIX

code --install-extension talktomebaby-0.3.1.vsix

Or via the UI: Cmd+Shift+P > Extensions: Install from VSIX... > select the file.

Build from source

git clone https://github.com/rishmadaan/talktomebaby.git
cd talktomebaby
npm install
npm run package   # produces talktomebaby-0.3.1.vsix
code --install-extension talktomebaby-0.3.1.vsix

Quick start

1. Open any .md, .txt, or other supported prose file
2. Cmd+Shift+P > TalkToMeBaby: Read Document
3. The Reader panel opens; reading starts from the beginning

To start from a specific point: place your cursor, then use TalkToMeBaby: Read from Here (right-click menu or command palette).

Reader panel

The Reader panel renders your document as a clean reading view. Each word is individually clickable.

• Click any word to jump playback there and resume from that word
• Scroll freely - auto-scroll follows playback while you don't interact; scroll away and a Return to playback pill appears in the corner; click it to snap back and re-engage auto-scroll
• The player bar at the bottom has speed presets, a fine slider, pause/resume, and stop
• The gear (⚙) at the right end of the player bar opens an in-reader settings panel for provider, voice, font size, and highlight colors (see Switching provider and voice)

Speed

Speed is set via the player bar in the Reader panel. Presets: 0.5, 0.75, 1.0, 1.25, 1.5, 1.75, 2.0x. The fine slider covers the full 0.5-2.0 range. Pitch is preserved at all speeds using preservesPitch. The chosen speed is persisted globally and restored on the next read.

Providers

Network providers synthesize audio by sending the text being read to the selected provider. Edge, ElevenLabs, and Sarvam require internet access; macOS say runs locally and does not send text to a network TTS provider. See PRIVACY.md for the full data-flow summary.

Switching provider and voice

The settings panel inside the Reader is the primary way to change everything. Click the gear (⚙) in the player bar to open it; click the gear again, press Esc, or hit the ✕ to close it. The gear shows a pressed state while the panel is open. The panel slides in above the player bar with two sections:

• Provider — one row per available provider, each with a short description, a "key required" badge where relevant, and a check (✓) on the active one. Click a row to switch. (macOS say only appears on macOS.)
• Voice — a dropdown of the active provider's voices, with the current one selected.
• Appearance
  • Font size — a +/- stepper (12-28px) that applies instantly and persists.
  • Highlight colors — sentence-band and current-word color pickers that apply instantly, plus Reset to theme to clear both back to the theme default.

(Speed lives in the player bar, not the settings panel.)

Instant open: the panel opens immediately. Voices are cached per provider for the session and prefetched when a read starts, so the first open is usually instant; if a provider's voices are still loading you'll briefly see a disabled "Loading voices…" option, then the list fills in.

Switching keeps your place and never auto-plays: changing provider or voice reconfigures the reader in place — the Reader is not torn down and rebuilt. The session re-primes paused at your current sentence with the new voice; press play when you're ready. No surprise audio. The settings panel stays open across the switch.

If a key-required provider has no stored key, switching to it prompts for the key first; cancel and the selection snaps back to the current provider.

Command palette (alternative, mirrors the gear panel):

• Switch provider: Cmd+Shift+P > TalkToMeBaby: Select TTS Provider — marks the active provider with ✓ / "(current)", and reconfigures in place on change.
• Select voice: Cmd+Shift+P > TalkToMeBaby: Select Voice — same active marker; reconfigures in place.
• Set API key: Cmd+Shift+P > TalkToMeBaby: Set API Key

Editor surface

While a session is active, the source editor stays in sync with the reader:

• Sentence decoration - the sentence currently being spoken is highlighted in the source editor
• alt+j (in the source editor) - jumps playback to the word at cursor; always starts playback even if paused
• editorClickToJump setting - controls whether a plain click in the editor also triggers a jump (see settings table below)
• Document edits - if you edit the file mid-read, playback pauses and a prompt offers to restart from the current position or stop

Settings

Privacy and API keys

TalkToMeBaby does not include analytics or telemetry. When you use a network TTS provider, the text being read is sent to that provider to generate audio. Provider keys (ElevenLabs, Sarvam) are stored via the VS Code SecretStorage API, which means your OS's encrypted credential store: Keychain on macOS, Credential Manager on Windows, libsecret on Linux. Keys are never written to settings.json, never synced, and never logged. See PRIVACY.md, DISCLAIMER.md, and the provider architecture notes for the full picture, including honest fragility notes per provider and the roadmap for a fully offline, OS-agnostic voice engine.

Keybindings

Commands

Troubleshooting

Edge TTS produces no audio / connection error Edge TTS requires an internet connection to Microsoft's speech service and is an unofficial, best-effort provider. If you are offline, switch to macOS say (talktomebaby.provider: say) or a cached document.

macOS say not available The say provider only works on macOS. On other platforms it is hidden from the provider picker.

ElevenLabs returns 401 ElevenLabs free accounts may be blocked from the TTS API. A paid plan (Starter tier or higher) is recommended for API access.

First read of a long document is slow Audio is synthesized in chunks as you read - subsequent sentences are prefetched in the background. Once a chunk is cached to disk, replaying it is instant and costs no API credits.

Playback highlight is off after editing If you edit the document while reading, the session model becomes stale. Use the prompt that appears to restart from your current position.

Playback stopped after a VS Code reload Reloading the VS Code window or restarting the extension host always ends the active session — audio is not preserved across reloads. If TalkToMeBaby saved your position (within the last 3 seconds of playback), you will be offered a "Resume?" toast on the next activation.

Reliability

Resume after reload: If VS Code reloads (window reload, extension host restart, OS sleep/wake), TalkToMeBaby saves your position automatically (every ~3 seconds while playing). When the extension reactivates, it offers a "Resume?" toast if a session was saved within the last 12 hours. Choosing Resume re-opens the document and jumps straight to your last sentence. Dismissing clears the saved position. The position is also cleared when you press Stop or when a document finishes naturally.

Note: a reload kills the active audio session — there is no way to resume mid-sentence. Resume starts from the beginning of the sentence you were on.

Contributing and security

Contributions are welcome; see CONTRIBUTING.md for local setup, test commands, and architectural invariants to preserve. Please follow the Code of Conduct.

For vulnerability reports or sensitive security issues, see SECURITY.md. Do not open public issues that contain API keys, tokens, or private document contents.

License

MIT. Additional disclaimers are in DISCLAIMER.md. Third-party runtime dependency notices are in THIRD_PARTY_NOTICES.md.