Opera 0.9.15 gives the assistant a voice and brings it to the desktop. It also lets you edit what you wrote and shape how your Maestro butterfly looks.
A voice for Opera
- An optional offline voice pack based on Supertonic speaks replies as they are written. Without it, Opera falls back to system speech.
- Each butterfly maps to its own Supertonic voice.
- Onboarding has a startup sound and a voice step.
- Voice settings unlock once the pack is installed.
Desktop
- The desktop layout was reworked and no longer follows the mobile port.
- A desktop voice assistant with an always-on clock screen is the start of the stand-by experience.
- A screen saver with its own settings, sleep dimming and OLED anti burn-in.
Edit your messages
You can edit a message you sent. Answers from the AI are gated, so an edit only applies where it makes sense. On Android, the back button cancels edit mode.
Video attachments and model settings
- Attach videos, with audio and video sniffing for OpenAI-compatible servers.
- Override the capabilities of each model in settings.
- Turn individual Ollama and cloud API models on or off.
- Cloud API models open in a sheet with a page per model, and the cloud provider was renamed to Cloud API.
- Built-in models are merged into on-device, and Maestro has its own settings page with a card.
Your butterfly
The Maestro butterfly flies from the home screen into the chat, now on Reanimated. You can customise it, and the profile card uses dynamic textures.
Reliability
- The overlay lifecycle is hardened, and background work pauses while it is open.
- The service pair is persisted atomically, and long model lists are virtualised.
- Late thought chunks are ignored and unreadable tool calls are reported.
- Android no longer asks for broad media permissions, to follow the Play photo picker policy.
- A debug build variant is available for Android installs.
Opera is open source. Follow the work on GitHub.