Turn any text into a beautifully narrated audiobook — right from your desktop, no GPU required.
Load a PDF, EPUB, plain text file, or paste content directly. Pick a voice (or clone your own from a short audio clip), adjust the speed and tone, and generate a full MP3 audiobook. The app intelligently splits your text at sentence boundaries, generates each section as an atomic sentence unit, and merges everything into one file.
Powered by PocketTTS by Kyutai — a lightweight text-to-speech model that runs entirely on CPU.
- Atomic Sentence Synthesis — Generates text sentence-by-sentence to prevent deep-learning hallucination and preserve speech quality.
- Line-Break Un-wrapper — Features an intelligent plain-text preprocessor that eliminates sentence fragmentation caused by hard-wrapped paragraph line breaks.
- 100-Word Safety Partitioning — Automatically partitions overly long sentences exceeding 100 words into natural atomic sub-sentences to maintain model stability.
- Self-Saving Hugging Face Token Popup — Prompts for and locally caches your Hugging Face Access Token when using voice cloning for the first time.
- Voice Cloning — Clone any voice from a short
.wavor.mp3sample. - 8 Built-in Voices — alba, marius, javert, jean, fantine, cosette, eponine, azelma.
- Import Anything — PDFs, EPUBs, TXT files, or scrape text from any URL.
- WPS Telemetry — Displays real-time words-per-second (WPS) speed and average processing times.
- Tone & Speed Control — Adjust expressiveness and playback speed.
- Auto MP3 Export — Merges all sentences into a single MP3 file.
- Stop & Resume — Pause generation and pick up precisely from any sentence index.
- No GPU Needed — Runs entirely on CPU.
- One-Click Install — Standard virtual environment
.venvlauncher handles everything automatically.
Double-click run_AudioStudio_pocketTTS.bat — it creates a local Python virtual environment, installs the lightweight CPU PyTorch runtime, extracts FFmpeg locally, and boots up the Audiobook Studio.
Note
Hugging Face Gated weights: PocketTTS requires accepting kyutai's terms on their Hugging Face page to download the voice cloning model. The app will automatically prompt you with an interactive dialog to enter and locally save your access token when cloning a voice for the first time.
If you prefer to run it manually from a terminal:
python -m venv .venv
.venv\Scripts\activate
pip install torch --index-url https://download.pytorch.org/whl/cpu
pip install -r requirements.txt
python PocketTTSUI.py| Button | What it does |
|---|---|
| Browse (Ref Audio) | Select a 5–10s audio clip to clone that voice |
| Browse (Output Dir) | Choose where your audiobook WAVs and MP3 are saved |
| 📄 Load PDF/Text/EPUB | Import text from a document |
| 🌐 Scrape from URL | Pull article or chapter text from a web page |
| Export Sentence | Save the current sentence .txt file for review |
| 📁 Open Folder | Open the output directory |
| ▶ Generate Speech | Start generating your audiobook |
| ⏹ Stop | Cancel safely after the current sentence finishes |
| 🔊 Quick Sample | Preview your voice + settings before committing |
| Setting | What it controls |
|---|---|
| Voice Name | Gated voice profile (used when no audio clip is provided) |
| Temperature | Lower (0.3 - 0.5) = consistent narration, Higher = expressive |
| Speed | 0.5x to 2.0x playback speed |
| Start Sentence | Resume precisely from any sentence index |
| Combine into MP3 | Merge all WAV sections into a single MP3 file |
- Temperature: 0.3–0.5 for stable audiobook narration, 0.8+ for dramatic reads.
- Interrupted? Set "Start Sentence" to resume exactly where you left off.
- Saved Files: Temporary output files generate as
output_1.wav,output_2.wav, etc., which are merged into a single MP3 (final_output.mp3or your custom name) on completion.
Please use this tool responsibly.
Voice cloning technology is extremely powerful but carries ethical risks.
- Do not clone voices without consent.
- Do not generate content intended to deceive, defraud, or harass.
- Always label AI-generated content appropriately.