StorySmith is an action, indie, RPG, and adventure game developed for PC that turns a standard streaming session into an interactive narrative experience. Viewers shape the direction through votes while an AI handles the writing and unexpected results, creating self-contained stories that conclude with a clear ending rather than endless continuation.
Gameplay
The core loop begins when the streamer defines the starting premise, including the genre, main character, and any desired twists. The AI then generates the opening scene based on that setup. From there, the audience participates by entering commands like !1, !2, or !3 to select from presented options. Vote counts remain hidden during the timer period to prevent early momentum from influencing late voters. Once the timer ends, the winning choice locks in, and the AI determines the actual outcome, which can include surprises such as alarms triggering or clean successes turning into complications.
Every decision carries genuine uncertainty because chat selects only the intended action while the AI controls the consequences. This separation keeps the experience fresh even for the streamer who reads the results live. Runs stay focused on reaching a defined conclusion instead of extending indefinitely, and the same initial premise can produce entirely different paths across multiple sessions due to varying chat input and AI responses.
Technical flexibility supports different setups. The game defaults to a local AI model that runs entirely on the user's machine with no recurring costs. Users who prefer alternative quality levels can connect their own Anthropic, OpenAI, or Google API keys instead. A resume function allows continuation of an interrupted run if the stream or computer restarts mid-session.
Game Modes
Three difficulty settings adjust the stakes and recovery options while remaining integrated into the story itself. Hard mode treats any death as a permanent end to that run. Medium mode provides a limited number of lives, each resolved through a narrative recovery scene such as a healer visit or cloning chamber that fits the chosen genre. Easy mode incorporates story-based narration to bypass trouble and continue without penalty.
These options affect how consequences play out rather than applying generic stat changes. The recovery moments in medium difficulty emerge naturally from the world and premise selected at the start, maintaining consistency with the ongoing narrative.
Streaming Integration
Built specifically around live Twitch interaction, the system supports one vote per viewer and processes commands in real time. Hidden tallies during voting encourage independent choices instead of bandwagon effects. The AI then generates scenes and consequences that reflect the selected action while introducing variability that even the streamer cannot fully predict in advance.
Each session produces unique results from identical starting conditions because chat composition and AI generation differ every time. This repeatability supports multiple streams with the same premise without repetition in the actual events or outcomes.
Is It Worth Playing?
StorySmith suits streamers who want audience participation to drive narrative direction while retaining unpredictability in results. The combination of chat voting, AI-authored consequences, and adjustable difficulty creates sessions that feel distinct across runs. Offline operation on a bundled local model removes subscription barriers, and the option to use personal API keys provides flexibility for those seeking different generation styles.
Players who enjoy choose-your-own-adventure formats with real risk and story integration will find the mechanics align closely with that preference. The self-contained arcs and built-in recovery systems keep focus on progression toward endings rather than prolonged grinding. Since the game emphasizes live interaction, it delivers the most value in a streaming context where chat engagement forms the central loop.
- Live chat voting with hidden tallies until reveal
- AI scenes and consequences that respond to but do not guarantee chat intentions
- Difficulty levels with fiction-integrated death and recovery systems
- Local model support for offline use or bring-your-own-key cloud options
- Resume capability for interrupted sessions
