Starstruk Logo MarkStarstruk

v1.5.0

FranckyB/Voice-Clone-Studiov1.5.0Feb 10, 2026by FranckyB

AI Summary

Feature release introducing Prompt Manager for local LLM-powered prompt generation using llama.cpp, standalone emotion storage, and new Sound Effects tool using MMAudio for text-to-audio and video-to-audio generation.

Key Highlights

  • Prompt Manager tool with llama.cpp integration
  • Sound Effects (MMAudio) tool
  • Standalone emotions.json storage
  • Local LLM generation with Qwen3 models
  • Custom model support for both LLM and audio

New Features

  • Prompt Manager with save, delete, clear functionality
  • Prompts stored in local prompts.json
  • LLM Generation using Qwen3-4B or Qwen3-8B GGUF via llama.cpp
  • System Prompt Presets for TTS/Voice and Sound Design/SFX
  • Model Auto-Download from HuggingFace
  • Custom Models - drop .gguf files into models/llama/
  • Automatic llama.cpp server management
  • Standalone emotions.json (migrated from config.json)
  • Clear VRAM stops LLM server
  • Sound Effects tool using MMAudio (CVPR 2025)
  • Text-to-Audio generation
  • Video-to-Audio generation
  • Multiple Model Sizes (Medium 2.4GB, Large v2 3.9GB)
  • Custom MMAudio model support (.pth or .safetensors)
  • Video Preview toggle

Full Release Notes

#### Version 1.5.0 - Prompt Manager & Emotion Storage

**Prompt Manager**

- **New Prompt Manager Tool** - Save, browse, and generate text prompts with a built-in LLM generator powered by llama.cpp
- **Saved Prompts** - Store prompts in a local prompts.json file with save, delete, and clear functionality
- **LLM Generation** - Generate prompts locally using Qwen3-4B or Qwen3-8B GGUF models via llama.cpp (no cloud API)
- **System Prompt Presets** - Built-in presets for TTS/Voice and Sound Design/SFX workflows, plus a custom option
- **Model Auto-Download** - Download Qwen3 models directly from HuggingFace into models/llama/
- **Custom Models** - Drop any .gguf file into models/llama/ to use your own models
- **Automatic Server Management** - llama.cpp server starts/stops automatically, cleaned up on exit or Clear VRAM

**Standalone Emotion Storage**

- Standalone emotions.json - Emotion presets are now stored in a dedicated emotions.json file instead of inside config.json
- Automatic Migration - Existing emotions in config.json are automatically migrated to the new file on first launch
- Independent Reset - Resetting config.json no longer wipes saved emotion presets

**Quality of Life**

- Clear VRAM Stops LLM - The Clear VRAM button now also shuts down the llama.cpp server if running
- SFX Filename Simplification - Sound effect filenames now use the first 8 words of the prompt instead of 40-char truncation with timestamp

**Sound Effects (MMAudio)**

- New Sound Effects Tool - Generate sound effects and ambient audio using MMAudio (CVPR 2025, MIT license), supporting both text-to-audio and video-to-audio modes
- Text-to-Audio - Describe any sound and generate 44.1kHz audio with adjustable duration, guidance strength, and negative prompts
- Video-to-Audio - Drop in a video clip and MMAudio generates matching sound effects synchronized to the visual content
- Multiple Model Sizes - Choose between Medium (2.4GB) and Large v2 (3.9GB) built-in models, with support for custom models
- Custom Model Support - Load your own .pth or .safetensors MMAudio checkpoints with automatic architecture detection
- Video Preview - Source/Result toggle to compare original video against the generated audio-muxed result