Skip to content

Latest commit

 

History

History
298 lines (193 loc) · 8.98 KB

ROADMAP.md

File metadata and controls

298 lines (193 loc) · 8.98 KB

🗺️ Velloris Roadmap

This document outlines the planned features and improvements for Velloris.

Current Version: v2.0.0

Released: February 2026

Major Features:

  • ✅ Three-mode architecture (realtime, dubbing, creative)
  • ✅ Proper PersonaPlex-7B integration (end-to-end S2S)
  • ✅ Ollama made optional (only for creative mode)
  • ✅ 10-15x performance improvement in realtime mode
  • ✅ Voice cloning support (dubbing/creative modes)
  • ✅ Comprehensive documentation

See CHANGELOG.md for full release notes.

Near-Term (v2.1 - Q1 2026)

Performance Optimizations

Priority: High

  • Model quantization support

    • Goal: 8-bit and 4-bit quantization for PersonaPlex-7B.
    • Outcome: Reduce VRAM requirements from 16GB to 8GB with a latency increase of no more than 10%.

  • Streaming TTS in dubbing mode

    • Goal: Generate audio progressively instead of waiting for full synthesis.
    • Outcome: Reduce perceived latency for long scripts by 50% and enable real-time playback during generation.

  • Batch processing optimization

    • Goal: Parallel processing for multiple scripts.
    • Outcome: Smart batching to maximize GPU utilization, with progress tracking and cancellation support.

Target: v2.1.0 release by April 2026

Audio Quality Improvements

Priority: Medium

  • Audio post-processing pipeline

    • Goal: Implement a post-processing pipeline with noise reduction, normalization, compression, and EQ adjustment.
    • Outcome: Achieve broadcast-quality audio output.

  • Voice mixing and effects

    • Goal: Add support for voice mixing and effects like reverb, echo, chorus, pitch shifting, speed adjustment, and background music mixing.
    • Outcome: Users can create more dynamic and engaging audio content.

  • Multi-speaker support

    • Goal: Detect speaker changes in script and automatically assign different voices.
    • Outcome: Maintain speaker consistency in multi-speaker scripts.

Target: v2.1.1 release by May 2026

Mid-Term (v2.2-v2.3 - Q2-Q3 2026)

Language Expansion

Priority: High

  • Enhanced multilingual support

    • Goal: Improve quality for non-English languages, with a focus on French, German, and Spanish.
    • Outcome: Achieve human-level quality for the target languages.
    • Dependencies: Language-specific voice models.

  • Dialect support

    • Goal: Add support for regional accents and cultural pronunciation nuances.
    • Outcome: User-selectable dialect variants for English (US, UK, Australian), Spanish (Spain, Mexico), and French (France, Canada).

Target: v2.2.0 release by July 2026

API & Integration

Priority: High

  • RESTful API server

    • Goal: Implement a RESTful API server with endpoints for all three modes.
    • Outcome: Enable easy integration with other applications and services.

  • Python SDK improvements

    • Goal: Improve the Python SDK with type hints, async/await support, context managers, and better error messages.
    • Outcome: A more developer-friendly and robust SDK.

  • Streaming API

    • Goal: Implement a streaming API using Server-Sent Events (SSE) and WebSockets.
    • Outcome: Enable real-time, low-latency audio streaming.

Target: v2.2.1 release by August 2026

Platform Support

Priority: Medium

  • Docker container

    • Goal: Provide an official Docker image with multi-architecture support (AMD64, ARM64).
    • Outcome: Simplify deployment and ensure consistency across different environments.

  • Cloud deployment guides

    • Goal: Create deployment guides for popular cloud platforms (AWS, GCP, Azure) and services (RunPod, Vast.ai, Lambda Labs).
    • Outcome: Lower the barrier to entry for cloud-based deployments.

  • Mobile support (experimental)

    • Goal: Develop experimental iOS and Android apps with on-device inference.
    • Outcome: Showcase the potential of Velloris on mobile devices.
    • Dependencies: Model quantization support (v2.1).

Target: v2.3.0 release by September 2026

Long-Term (v3.0+ - Q4 2026 and beyond)

Advanced Features

Priority: Medium

  • Emotion detection and matching

    • Goal: Detect emotion from user voice in realtime mode and generate a response with matching emotion.
    • Outcome: More natural and engaging conversations.

  • Voice conversion

    • Goal: Implement real-time voice conversion, including cross-gender and age progression/regression.
    • Outcome: A powerful tool for content creators and privacy-conscious users.

  • Speech-to-speech translation

    • Goal: Translate speech while preserving voice characteristics in a multi-language conversation mode.
    • Outcome: Break down language barriers in real-time communication.

Target: v3.0.0 release by Q4 2026

Model Improvements

Priority: High

  • Fine-tuning support

    • Goal: Add support for fine-tuning PersonaPlex-7B and Qwen3-TTS for custom voices and specialized domains.
    • Outcome: Users can create their own high-quality custom voices.

  • Custom model support

    • Goal: Implement a plugin system for third-party models, including new TTS engines and LLMs.
    • Outcome: A more flexible and extensible platform.

  • Zero-shot voice cloning

    • Goal: Improve voice cloning with minimal reference audio (e.g., a single sentence).
    • Outcome: Make voice cloning more accessible and easier to use.

Target: v3.1.0 release by Q1 2027

Enterprise Features

Priority: Low (community-driven)

  • Multi-tenancy support
  • Monitoring and observability
  • High availability

Target: v3.2.0 release by Q2 2027

Research & Experimental

Priority: Low (exploratory)

These features are under investigation and may or may not be implemented:

Advanced Reasoning

  • Chain-of-thought speech
  • Multi-turn planning

Audio Understanding

  • Speaker diarization
  • Acoustic scene analysis

Multimodal Support

  • Vision integration
  • Text + audio input

Community Requests

Features requested by the community. Vote on GitHub Discussions. The top 3 most upvoted features will be considered for the next release cycle.

High Demand

  1. Web UI / GUI (150+ votes)
  2. More voice options (120+ votes)
  3. Voice editor (100+ votes)

Medium Demand

  1. Plugin system (75+ votes)
  2. Mobile apps (60+ votes)
  3. Video dubbing (50+ votes)

How to Contribute

We welcome contributions to any feature on this roadmap!

For Developers

  1. Check open issues labeled help wanted
  2. Comment on the issue to claim it
  3. Fork, implement, and submit a PR
  4. See CONTRIBUTING.md for guidelines

For Researchers

  1. Open a discussion in GitHub Discussions
  2. Share your research findings or ideas
  3. Collaborate on experimental features
  4. Co-author papers on Velloris innovations

For Users

  1. Vote on features in GitHub Discussions
  2. Share your use cases and requirements
  3. Test beta features and provide feedback
  4. Report bugs and suggest improvements

Versioning Strategy

Velloris follows Semantic Versioning:

  • Major versions (v2.0, v3.0): Breaking changes, major new features
  • Minor versions (v2.1, v2.2): New features, backward compatible
  • Patch versions (v2.0.1, v2.0.2): Bug fixes, backward compatible

Release Cadence

  • Major releases: 6-12 months
  • Minor releases: 1-3 months
  • Patch releases: As needed (critical bugs)

Completed Features (v2.0)

Features from the original v1.x roadmap that are now complete:

  • ✅ Three-mode architecture (v2.0.0)
  • ✅ Proper PersonaPlex usage (v2.0.0)
  • ✅ Optional Ollama (v2.0.0)
  • ✅ Voice cloning (v2.0.0)
  • ✅ Comprehensive documentation (v2.0.0)

Deprecated Features

Features from v1.x that are no longer supported:

  • ❌ Interactive mode (deprecated in v2.0)

Version History

Version Release Date Status Highlights
v2.0.0 Feb 2026 Current Three-mode architecture, proper PersonaPlex usage
v1.0.0 Jan 2026 Deprecated Initial release, interactive mode

Feedback & Suggestions

Have ideas for the roadmap? We'd love to hear from you!

Your feedback shapes the future of Velloris!

Last updated: February 2026

For the latest updates, see CHANGELOG.md.