Skip to content

[Feature Request]: Agentic Multi-Speaker "Audio Overview" (Podcast) Generation from Knowledge Base #613

Description

@MFXM

Do you need to file a feature request?

  • I have searched the existing feature request and this feature request is not already filed.
  • I believe this is a legitimate feature request, not just a question or bug.

Feature Request Description

Description

DeepTutor excels at transforming static knowledge bases into dynamic interactions via Deep Research, Deep Solve, and Quiz Generation. However, it currently lacks an auditory learning dimension. Implementing an automated "Audio Overview" generation pipeline—similar to Google NotebookLM / Illuminate—would be a massive milestone for multi-modal learning.

Since the repository already contains structural endpoints for TTS providers that are currently underutilized, extending the system to support agent-driven conversational audio generation is a logical evolutionary step.

Proposed Solution

  1. Dialogue Orchestration: Introduce a new agentic pipeline option (Audio Overview). This triggers a dual-agent loop where two distinct system personas (e.g., Host A: The Inquisitive Reviewer and Host B: The Domain Expert) synthesize a structured, conversational script mapping out the macro themes, key citations, and core equations of an indexed document.
  2. Multi-Speaker TTS Mapping: Parse the generated script blocks and feed them sequentially into the configured TTS backend. The system should map the separate agent IDs to distinct voice profile variables (e.g., using alternative voice seeds via OpenAI TTS, ElevenLabs, or local lightweight runtimes like Kokoro).
  3. Artifact Caching: Save the compiled audio file (.mp3 or .wav) alongside the generated markdown transcript inside the workspace artifact layer (data/user/artifacts/audio/), allowing users to play, scrub, or download the digest straight from the Web UI.

Related Module

Guided Learning

Use Case

  • True Multi-Modal Learning: Dramatically improves accessibility for students, scientists, and auditory learners who need to consume heavy literature or lecture notes during commutes or away from screens.
  • Exploiting Existing Hooks: Unlocks the latent capability of the current TTS configurations, moving the codebase from primitive line-reading to fully orchestrated multimedia generation.
  • Academic Science Communication: Simplifies public outreach by giving doctoral researchers and educators an automated way to spin up high-fidelity audio overviews of their newly indexed preprints or group publications.

Additional Context

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Fields

    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions