Charisma

Author conversations with a purpose

Charisma is a platform for building interactive conversational experiences, including training scenarios, character interactions and brand campaigns. It combines authored dialogue and branching story logic with AI interpretation and optional generated replies. A learning designer or narrative team can define what a character knows, how a conversation progresses and which choices influence the next response.

It is relevant when a conversation needs an intentional structure rather than an unrestricted chat session. For example, a customer-service exercise can let a learner respond in their own words while keeping the scenario focused on a particular situation and learning objective.

A graph of dialogue, responses and memory

The Quick Start describes a story graph made from connected nodes. Character nodes supply dialogue, player nodes represent possible user responses, and memory nodes retain information for later use. The author writes the available routes; Charisma interprets a user's response and attempts to match it to the appropriate path.

Matching documentation explains why authors should cover different intentions and include wildcard routes for unexpected input. Memories and gates let later interactions depend on earlier decisions or saved information. A graph needs an available fallback when a condition is not met, otherwise the conversation can stop.

Mix authored and generated replies carefully

Generation nodes bring language-model replies into an authored experience. Prompts can use story context and memories, and generated text can feed a character response. This supports variation while retaining a planned conversation structure.

The documentation also sets practical boundaries: generation has an output-length limit, chained generation nodes add latency, and failed or moderated requests need another route. Write concise instructions and retain authored responses for moments where exact wording matters. A generated character response should not be treated as a verified statement of policy or a reliable assessment of a learner.

A customer-service practice scenario

Suppose a training team wants staff to practise responding to a delayed delivery. Create a customer character, an opening complaint and separate paths for acknowledging the problem, asking a useful question or offering an inappropriate promise. Save a decision when the learner clarifies the delivery details.

Use that memory to open a later resolution path, and write a fallback for answers outside the expected examples. Add limited generated dialogue only where variation helps the exercise. Run the conversation through the built-in chat tester using several different phrasings, then ask colleagues to try unexpected responses. Review where the intended lesson becomes unclear before integrating the experience into the learner-facing interface.

Deployment and review

The documentation describes SDK integration with external clients, including web applications, Unity and Unreal Engine. This means a published training or game experience can require interface and integration work alongside story authoring.

Published-version analytics include playthrough starts, finishes, conversation time, route usage and transcripts. These can help identify unmatched responses or confusing branches. Completion statistics describe interaction with the scenario; they are not automatic proof of learning or business impact.

Access and project requirements

Charisma's pricing page describes a paid Pro route using credits for SDK-based projects and an Enterprise offering with tailored development and platform fees. Account setup is contact-led. Confirm usage, voices, languages, integration support and data handling for the actual project. Budget for writing, review and implementation as well as runtime access.