Introducing WISP — Wrist Intelligence & Speech Portal
Formerly Gemini with Voice, WISP has evolved into a universal AI voice assistant for Pebble smartwatches.
What started as a voice interface for Google Gemini has become a flexible platform for interacting with multiple AI assistants, with independent control over text-to-speech providers.
What's New
- New identity: Rebranded from Gemini with Voice to WISP (Wrist Intelligence & Speech Portal).
- Universal AI support: Connect to compatible AI agents through installed Pebble phone plugins, including Gemini, ChatGPT, Claude, Grok, and other supported agents.
- Universal TTS integration: Choose from multiple speech synthesis providers through the Universal TTS plugin.
- Updated interface: New WISP branding, app icons, launcher icon, and provider-specific logos.
- Automatic agent recognition: Display the selected AI provider's name and icon when available.
- Improved audio reliability: Better Bluetooth buffering, packet acknowledgment, retry handling, and continuous speech playback.
- Legacy compatibility: Existing Direct Gemini API and Google Cloud TTS configurations remain supported.
- Personalization: Retains adjustable volume, text size, custom AI instructions, and automatic light/dark appearance.
Compatibility
WISP preserves the original Gemini with Voice application identity, allowing existing users to retain their settings when upgrading.
The rebrand introduces a more flexible architecture without requiring existing users to abandon their working configurations.
*Note: Plugin-based features require compatible plugins installed and configured in the Pebble phone app.*
# Gemini with Voice — Version 1.5.0
## What's New
Version 1.5.0 is a major update focused on smoother voice playback, improved reliability, expanded Pebble compatibility, and better error handling.
### Improved Voice Playback
- **Continuous audio playback:** Keeps the speaker amplifier active throughout an entire response, eliminating popping noises between sentences.
- **Smarter sentence buffering:** Each sentence now builds its own audio buffer before playback begins, reducing stuttering and interruptions.
- **Adaptive buffering:** Automatically adjusts buffering requirements based on Bluetooth transfer speed and sentence length.
- **Smoother audio processing:** Uses smaller 40 ms audio blocks for more consistent playback, particularly around commas and natural pauses.
- **Improved silence handling:** Maintains amplifier power using short silent audio frames without unnecessarily inserting silence into speech.
- **Clean audio shutdown:** Closes the speaker only after the complete response has finished, with a short silent tail.
- **Better interruption recovery:** Handles Bluetooth delays, audio underruns, and interrupted playback more reliably.
### Expanded Pebble Compatibility
Now supports six Pebble platforms:
- **Pebble Time 2:** Full voice and text support.
- **Pebble 2 Duo:** Full voice and text support.
- **Pebble Round 2:** Voice dictation with text responses.
- **Pebble Time / Time Steel:** Voice dictation with text responses.
- **Pebble Time Round:** Voice dictation with text responses.
- **Pebble 2:** Voice dictation with text responses.
Speaker playback is automatically enabled only on compatible hardware. Added appropriate display resources for the new platforms.
### Voice Selection Improvements
- Added 12 selectable Google Cloud Text-to-Speech voices across WaveNet, Neural2, and Chirp 3 HD.
- Added pricing tiers ($, $$, $$$) to make relative voice costs clearer.
- Removed expensive Studio voices from the selection.
- Fixed invalid Chirp 3 HD voice identifiers.
- Restored support for Neural2-F.
- Uses WaveNet-F as the default and fallback voice.
- Automatically migrates unsupported legacy voice selections.
### Reliability and Error Handling
- Added timeout protection for Gemini and Google Cloud Text-to-Speech requests.
- Improved handling of API errors, authentication failures, rate limits, and network interruptions.
- Strengthened Bluetooth audio transfers with packet acknowledgments, retries, and sequence verification.
- Prevents duplicate packets from corrupting audio playback.
- Improved cleanup of playback timers, audio buffers, and speaker resources.
- Prevents outdated requests from interfering with newer responses.
### Conversation History and Privacy
- Improved conversation history management, retaining up to three recent question-and-answer exchanges.
- Prevents failed requests from unnecessarily deleting valid conversation history.
- Added a **Clear Conversation History** option in Settings.
- Preserves API keys and other preferences when clearing history.
- Expanded privacy and API usage documentation.
### Testing and Build Improvements
- Added automated GitHub Actions regression testing.
- Added tests covering audio streaming, voice selection, settings, conversation history, and error handling.
- Added platform-specific speaker detection tests.
- Added automated Pebble SDK compilation for all six supported platforms.
- Updated the settings framework to the maintained `@rebble/clay` package.
- Corrected CloudPebble message-key configuration for the expanded audio streaming protocol.
- Synchronized application version information to 1.5.0.
## Summary
Gemini with Voice 1.5.0 delivers smoother, more natural voice responses, more reliable Bluetooth streaming, improved privacy controls, and broader Pebble compatibility.
**Compatibility note:** All six platforms compile successfully. Speaker playback has been physically tested on Pebble Time 2; Pebble 2 Duo and Round 2 hardware behavior has not yet been independently verified.
Removed old gemini models that have been depreciated. Fixed version options that stripped by accident in 1.4.3
Added a toggle to turn on and off TTS. TTS is no longer a requirement for the program
Leaving TTS API blank should skip the text to speech function so the app is easier to use and free since you don't need to pay $10 to google for the bridge (or whatever the call it).
Hoping a republish brings back the logo
Added Audio Quality Control: Introduced a new "Audio Stream Quality" configuration option in settings, allowing users to select between four different bit-depth and sample-rate profiles to balance sound fidelity against Bluetooth connection stability.
Default Stability Optimization: Reverted the default audio stream to 8kHz/16-bit to ensure maximum compatibility and glitch-free, gapless playback across all supported Pebble hardware.
Enhanced API Flexibility: Replaced hard-coded model and voice strings with dynamic dropdown menus in the settings page, providing seamless support for new Gemini model versions and high-fidelity TTS voice profiles.
SDK 4.3 Integration: Implemented quiet_time_is_active() support, allowing the application to automatically mute the speaker when the watch is in Quiet Time mode.
Improved Buffer Management: Refined the amplifier shutdown logic to prevent the audio thread from trapping the system in infinite countdown loops, significantly improving system stability during extended text reading sessions.
Optimized Button Mapping: Re-implemented the button configuration logic to ensure that audio volume adjustments via long-press do not conflict with native scroll-layer navigation, providing a more intuitive reading experience.
1.0 to 1.2.1 was designed specifically for the Pebble Time 2.
1.3 should be compatible with all Pebble units except for the OG Pebble and Pebble Steel as there is no microphone. For units without a speaker, you should only need the the Gemini API key.
Updated Settings Page
Provides feedback on API key stored status
API Setup guide at https://tinyurl.com/PebbleGemini
Audio enhancements
On main screen, up and down will change audio level in 10% increments with on display feedback
If set to 0%, no efforts will be made to download and play audio, saving on battery and bandwidth.
Adjusted audio stream to play 16bit compressed audio
Keeps amp on until audio stream ends, removes pops between buffering. Greatly increased audio quality.
Added graphics to the main screen and waiting screen
Added audio buffering notification