AI-assisted media creation
Placeholder media makes a prototype feel like the game. ChatMapper can generate a spoken take for a line, a portrait for a character, a picture for a place or a node, and a short video — from the text that is already in the project — and attach the result like any uploaded file.
Generated media is a workspace feature switch (Settings & Labs → Generative media and Generated actor portraits). It runs on provider keys the workspace admin adds under the AI tab; the included hosted allowance covers text generation only.
Spoken lines
Section titled “Spoken lines”On a node’s Media tab, the ✦ Generate with AI button on Audio files
sends the node’s dialogue to ElevenLabs and attaches the returned audio.
Sentence splits (|) become separate files, in order, pairing with the
sentences exactly as recorded audio would.
Each actor has a Voice (ElevenLabs) field in the Actors panel — paste a voice ID from your ElevenLabs library, and Preview voice to hear it. Every line that actor speaks then generates in that voice.
Markup is stripped before synthesis: [em1] tags, [var=…] tokens and stage
directions are not read aloud.
Actor portraits
Section titled “Actor portraits”In the Actors panel, Generate portrait builds a prompt from the actor’s name and description plus a project-wide style — one style shared by every actor and place, so the cast looks like it belongs in the same story — and asks OpenAI’s image model for a portrait. Preview it, then attach. The portrait then shows on node cards, in the simulator and in Play mode.
Actors without a portrait get a deterministic generated avatar automatically, so a cast is never faceless.
Place and node pictures
Section titled “Place and node pictures”The same Generate button exists on a Place (a scene backdrop for Play mode) and on a node’s Pictures slot. For a node, the generation direction is assembled from the story around it — who is speaking, where, what just happened — and you can edit it before generating, or let the ✦ art director rewrite it from that context.
Generate on a node’s Video file slot offers two modes:
- Avatar video — a talking-head render of the line via HeyGen or
Synthesia, using an avatar ID and voice ID you set per actor
(
ai_video_avatar_id,ai_video_voice_id). Renders are asynchronous; the slot polls and attaches the file when it is ready. - Scene video — a character-free cinematic clip from a prompt built from the node’s context, via OpenAI.
Where it goes
Section titled “Where it goes”Everything generated lands in the project’s media exactly like an upload: attached to the slot, stored once per workspace, included in CMPKG + media exports, replaced or detached whenever you like. Every generation is logged against the workspace.
Sensible expectations
Section titled “Sensible expectations”- Generated audio is a scratch track — right for pacing, timing and playtests; a recording studio will replace it, and the voiceover sheet is the hand-off.
- Portraits and pictures are concept placeholders. The project-wide style keeps them consistent enough to show a producer.
- Nothing is generated without you clicking Generate on a specific slot. There is no “generate all” that silently spends a key.