speak-gpt

• Added ability to lock assistant window. When assistant window is locked it can't be closed by swiping window down. Use "Exit" button instead. This feature can potentially save your conversation from accidently closing the assistant. This feature is optional and available as an experiment. If you don't see it, enable new settings UI by tapping "Lab" icon in settings menu.
• Other bugs fixed

[!NOTE]
Recently I owned new Poco device with OLED display and found that material you theme has broken on MIUI/HyperOS so I brought AMOLED mode for you. Please note that this feature is experimental. If you don't see this feature make sure you're using new settings layout. You can access it by tapping "Lab" icon in old settings page. AMOLED mode will work only if dark theme is enabled on your device.

• UI improvements
• Added full support for RTL locales
• Bugs fixed

• Desktop mode (autofocus keyboard, keyboard shortcuts: Enter to send, Shift+Enter to add new line)
• Various bugs fixed

Release 3.2 has been skipped for technical reasons

• Optimized for devices with physical keyboard
• Removed Markdown from user messages
• Various bugs fixed

• WSA and Google Chrome integration (View this repository for more info: https://github.com/AndraxDev/speakgpt-on-wsa)

• Added new cose effective model gpt-3.5-turbo-0125
• Chat autonaming is now using this model
• Removed deprecated models: ada, babbage, curie, davinci, other text models, gpt temporary models, etc.

• Bugs fixed

• Added autocapitalize flag on some fields
• Swithced to gpt-3.5-turbo-1106 model to generate chat names

[!NOTE]
Release 2.36 has been merged with 2.37 due to technical reasons.

• We're introducing a new OpenAI DALL-E 3 model.
• App has been redesigned. Interface now is more compact.

• OpenAI TTS

• Added GPT4 turbo model
• Added "auto" naming for chats
• Long press cancel button to discard recording
• If recorded audio does not contains any text it will be discarded
• All messages are now selectable (you can copy them)
• Autoscroll will be disabled if you scroll chat while generating text
• Summarization option is added to the assistant option menu
• A model name is displayed near each chat
• Added some additional tips to resolve confusions about settings
• Added German language
• Fixed bug when settings are not saved or lost
• Fixed system messages

• Now SpeakGPT will generate an image instead of text when clicking "Try it" button if prompt intended for DALL-e.

• Now you can search across your chats

• Extended support for older Android versions
• Other bugs fixed

• "explain meaning" feature is now managed by auto language detect.
• Added categories panel to the prompts store
• Localized some features
• App is now more user-friendly for disabled people

• "Explain text" feature is now managed by the "Assistant language" setting.

• Optimized for Android 14

Added action selector dialog when using assistant from context menu and share screen

• Added system messages