opentts

Project Overview

ForgedEcho is a native speech production app built for macOS that wraps several top open-source speech models in a graphical interface, sparing you from setting up Python environments and debugging command-line scripts. Cloning a voice, tuning emotion, dubbing subtitles in batch, turning a novel into an audiobook, or even training a custom voice model can all be done locally on one Mac.

From a five-second voice clone to a whole audiobook, everything stays local on your Mac.

Key Features

• Multiple engines: three built-in speech engines, switchable between fast generation, high-fidelity cloning and multilingual synthesis
• Voice cloning: clone a voice from a few seconds of reference audio, with reference fine-tuning for higher similarity
• Emotion control: multi-dimensional continuous emotion adjustment so one line can be delivered in different moods
• Dubbing studio: import a subtitle file, split by timeline automatically, assign voices per role and export a batch of synthesized audio
• Audiobook line: import a novel, split it into chapters, detect dialogue roles, resume from a breakpoint and export an audiobook with chapter navigation
• AI emotion director: analyzes the tone of each sentence and maps it to emotion parameters
• Voice lab: similarity comparison, blind A/B listening and a library-wide clash analysis
• Local LoRA training: slice audio, transcribe automatically and fine-tune a custom voice model on the Mac
• Quality loop: local speech recognition checks the result for missing or wrong characters and grades its quality
• Privacy vault: export voice libraries and project files as an encrypted backup package

Target Audience

• Content creators who need voice-over for videos
• Producers who want to turn novels into audiobooks
• Privacy-conscious users who want speech synthesis fully local

Quick Start

1. Clone the repository and run swift run ForgedEcho to build and launch
2. Open the settings page and choose the speech engine to use
3. Click the Start button on the toolbar; the app fetches models and the environment and warms up automatically
4. Back in the workspace, enter text, pick a voice and press Cmd + Return to synthesize and play
5. To produce a standalone app, run the bundled script to package a .app
6. For dubbing or audiobooks, import subtitles or text in the matching module and generate in batch

Core Advantages

• Fully offline: all inference runs locally, and no audio is uploaded
• No dependency setup: models and environment are prepared by the app instead of manual command-line work
• One production line: preview, dubbing, audiobooks, training and verification live in a single app
• Native experience: a pure native build that is small, starts fast and uses few resources

System Requirements

Item	Requirement
OS	macOS 14 or later
Hardware	Apple Silicon recommended, 16 GB of memory or more suggested
Build tools	A recent developer toolchain is needed when compiling from source

Notes

• The first launch downloads models and the environment, so leave room on disk and keep the network available
• Chinese synthesis and multilingual ability differ between engines, so pick the engine that suits your material
• Local training needs clean voice recordings, and sample quality directly affects the result
• Do not use the tool to produce content that infringes others' rights or violates the law