Saga Studio

Saga Studio turns a manuscript into a story you can sit inside: voiced, illustrated, sometimes branching. It runs locally, on your own machine, and nothing leaves it. Drop a markdown file in and it becomes a deck of cards: edit a line, re-render just that line, listen, fix, move on. When it sounds right, publish it as an audiobook, an animatic video, or an offline web story, then share that web story to darkride.ai with one button.

What's in the box

• Voices from a whisper of audio. A 5–10 second reference clip is the whole setup for a cloned voice. No training step.
• Cards, not timelines. Every chunk's audio is content-addressed, so changing one word re-renders one card, not the book. Fixing a mispronunciation costs thirty seconds, not an evening.
• Music and effects. Audio cards fade and duck under the narration; voiced cards re-speak your own performance in a character's voice, keeping your timing and delivery.
• Pictures and choices. Visual cards put art on the stage; choice cards branch the story with flags and counters. The same story graph plays identically in the studio and in the exported player.
• Three ways out. Assemble an audiobook (mp3/m4a/flac/wav), export an animatic mp4, or export an offline web story that plays from a plain folder. The publish tab's Share button puts that web story on darkride.ai and hands you the link; sharing again updates the story and keeps the link.

Get it

Saga Studio is open source, built for macOS on Apple Silicon (CUDA works too). You need python3.11, ffmpeg (brew install ffmpeg), and the GitHub repo:

git clone https://github.com/caitlynmeeks/saga-studio
cd saga-studio
python3.11 -m venv .venv
./.venv/bin/pip install -r requirements.txt
./.venv/bin/python studio.py        # opens at http://127.0.0.1:5010

There is also a desktop app with a bundled runtime and the Kokoro voice engine built in; bigger engines download themselves from the Voices tab on first use. A packaged download is coming; until then the app builds from the saga-studio-electron repo with npm install && npm start.

Use it

Drop a .md file on the Stories tab and it becomes a project. Put a short reference clip in voices/ and it becomes a narrator. Edit cards, press render, listen in the Stage window, and when the story is ready, open the Publish tab and choose a door out. The README is the deep guide: voices and delivery controls, audio card fades, choice grammar, takes, and the discuss agent.

Voice engines and their licenses: Chatterbox (MIT), Kokoro (Apache-2.0), OmniVoice (CC-BY-NC, non-commercial). Your stories are yours; what you share on darkride.ai you can update or remove with the token your studio keeps.