murtaza-nasir/speakr v0.3.1-alpha

Simplified Setup, Speaker Diarization, and More!

on GitHub

This release introduces a host of new features and improvements, focusing on a more streamlined setup process, powerful speaker identification capabilities, and enhanced customization.

⚠️ Important Update Notice

This update includes changes to the docker-compose.yml file and introduces a new .env file for configuration. These changes may require manual updates to your existing setup.

Before updating, it is strongly recommended to back up your instance and uploads directories.

To update your installation, please follow the new setup instructions in the README.md file carefully. You will need to create a .env file and update your docker-compose.yml to match the new format.

Key Features

• Simplified Docker Setup: Configuration is now managed through a single .env file, making it much easier to get started and manage your settings. We've provided two templates: env.whisper.example for standard API-based transcription and env.asr.example for using an ASR webservice.
• Speaker Diarization: The application can now automatically identify and separate different speakers in a recording. This feature is enabled when using the ASR webservice method with the whisperx engine. Speakers are assigned generic labels (e.g., SPEAKER 01), which you can then easily name.
• Intuitive Speaker Identification: A new modal allows you to easily assign names to the detected speakers. The transcript is displayed with the current speaker highlighted to provide context.
• AI-Assisted Speaker Naming: Use the "Auto Identify" feature to let an AI model attempt to name the speakers for you based on the content of the conversation.
• Saved Speaker Profiles: You can now save identified speakers. These will be suggested for auto-completion in future transcriptions, saving you time. You can view and manage your saved speakers on your Account page.
• Transcription Reprocessing: A new "Reprocess" button lets you re-run a transcription with different settings, such as adding speaker diarization to a recording that was initially transcribed without it.
• Enhanced Summarization: Summaries are now more context-aware, incorporating your name and professional title. You can also provide a custom prompt to guide the summary generation process.