🚀 ChloeOS YTVoice: Cloud-Hosted Concatenative TTS Playhead Database
This video contains the concatenative playhead database for the ChloeVoice speech synthesizer, consisting of 13,126 raw voice sprites stitched together. Each word is separated by a 250ms silence guard band to prevent audio bleeding during playhead seeks.
By utilizing the companion subtitle (.srt) track, any system with internet access can use this video as a remotely hosted voice synthesizer—allowing low-resource devices to speak arbitrary sentences with 0% local audio storage constraints!
Inspired by how the Autobot Bumblebee speaks—scanning radio frequencies and stitching recorded voice fragments together on-the-fly—YTVoice treats this video as a remote soundboard.
—————————————————————-
🛠️ HOW TO USE: PLAYHEAD ARCHITECTURE
—————————————————————-
1. Load the SRT Map: Parse the companion .srt subtitle file into a light index mapping each word to its start timestamp and duration (in milliseconds).
2. Seek and Play: Use an on-the-fly slice tool to jump directly to a word’s start timestamp, play for its specific duration, and pause.
3. Sequence: Chain seeks together to stitch words dynamically on-the-fly.
—————————————————————-
⚠️ SYSTEM COMPATIBILITY NOTICE
—————————————————————-
✅ COMPATIBLE WITH: Linux (Ubuntu/Debian), macOS, Windows, Raspberry Pi, and Android Termux.
❌ NOT NATIVELY COMPATIBLE WITH: Standalone microcontrollers (ESP32, ESP8266, Arduino) due to full Python framework, FFmpeg, and yt-dlp dependencies.
—————————————————————-
📦 PREREQUISITES
—————————————————————-
Core Binaries Needed in System PATH:
• ffmpeg (For rapid millisecond audio slice extractions)
• mpv or ffplay (For automatic hardware media playback)
Python Ecosystem:
• pydub = 0.25.1
• yt-dlp = 2023.01.01
• soundfile numpy (Required only for offline binary compiler mode)
• audioop-lts (Conditional backport required for Python 3.13+)
—————————————————————-
🚀 CORE MODES OF OPERATION (Via Client Repository)
—————————————————————-
1️⃣ YouTube Cloud Mode: Bypasses local storage. Resolves this video stream using HTTP range requests to download only the precise word fractions needed on-the-fly.
👉 python3 src/player.py “your sentence text here” –youtube r-WQt6Hi86Y
2️⃣ Local Offline Mode: Slices a local media container file on disk instead of loading it entirely into RAM, keeping the memory footprint under 15MB.
👉 python3 src/player.py “your sentence text here” –audio database_speech.mp4
3️⃣ Compiled Binary Mode: Direct disk byte seeks against an uncompressed flat raw package with 0ms seek latency.
👉 python3 src/player.py “your sentence text here” –bin voice_sprites.bin –index voice_sprites.bin.index.json
—————————————————————-
🔗 SOURCE CODE
—————————————————————-
GitHub Repository: https://github.com/acadiemedia/ytvoice
Note: This project is an open-source project released under the MIT License. It is currently built as a proof-of-concept prototype.
#Python #TextToSpeech #FFmpeg #Ytdlp #Termux #OpenSource #RaspberryPi #Hacking #Transformers
source
