💡 Voice-Crafted Slide Decks: Avatar-Guided Presentation Creation as Competitive Response

[github-actions[bot]]

Summary

The avatar walks non-technical users through presentation creation conversationally: "What's your key message? Who's the audience? How many slides do you need?" Then generates, previews, and iterates on slides in the output panel — with export to PDF and PowerPoint. This is a direct competitive response to Claude Cowork's September 2026 presentation creation feature, differentiated by TalkTerm's conversational, avatar-guided approach.

Market Signal

Claude Cowork shipped dedicated presentation and document creation features in September 2026, with PDF and PowerPoint export (TechCrunch, Sept 16 2026). ChatGPT has Canvas for collaborative document editing. Knowledge workers spend an estimated 20% of their time on presentation creation. For TalkTerm's target non-technical users, the ability to go from idea to shareable presentation without leaving the application is becoming table-stakes functionality. The competitive pressure is acute: Cowork merged Chat/Code/Cowork into a single interface on September 16, 2026 — just days ago — with auto-routing between modes.

User Signal

PRD Journey 1: Sarah the PM produces a brainstorming artifact and shares it with her team. In practice, "sharing with the team" means a presentation — not a markdown document. If Sarah has to leave TalkTerm, open Google Slides or PowerPoint, and manually recreate the brainstorming output as slides, the "complete workflow" promise is broken. The output panel already supports multiple display modes (FR45: document, comparison table, clustered cards, activity feed) — slides are a natural and architecturally consistent extension.

Technical Opportunity

TalkTerm's architecture includes a multi-mode output display panel (FR45) rendered in Electron's Chromium renderer — an environment that already runs web content natively. A slides display mode using Reveal.js (MIT-licensed, HTML-based presentations, ~250KB gzipped) is architecturally minimal. Export paths:

• PDF: html2pdf or Electron's built-in webContents.printToPDF()
• PowerPoint: pptxgenjs (JavaScript PPTX generation library, MIT-licensed)

The avatar's conversational guidance is the key differentiator vs. Cowork's form-based approach. Users speak their intent ("Make a 5-slide deck about our Q3 results for the board"), the avatar asks clarifying questions about audience and key messages, generates slides, narrates the deck back for review, and iterates conversationally on feedback.

Assessment

Dimension	Score	Rationale
Feasibility	high	Reveal.js + pptxgenjs are mature JS libraries; output panel architecture supports new display modes
Impact	high	Addresses core user need (sharing work); matches competitor capability with voice-first differentiation
Urgency	high	Claude Cowork shipped this Sept 2026 — gap grows with each week of delay

Adversarial Review

Strongest objection: Chasing Claude Cowork's feature list dilutes TalkTerm's focus. Presentation creation is a commodity feature — better to nail the unique avatar + voice differentiator first. Adding Reveal.js + pptxgenjs increases dependency weight in an Electron app that should stay lean.

Rebuttal: This is not feature-chasing — it is closing a workflow gap. If Sarah produces a brainstorming document but cannot turn it into a presentation without leaving TalkTerm, the "complete workflow" promise breaks and she goes back to doing things manually. The differentiator is not the slide creation itself but HOW it happens: conversationally, with the avatar guiding structure, suggesting layouts, and narrating the deck back for review. Reveal.js adds ~250KB gzipped — trivial for an Electron app already shipping Chromium (~150MB). PPTX export can be lazy-loaded on first use.

Suggested Next Step

Add a slides display mode to the FR45 output panel specification. Define the avatar's presentation creation conversation flow: (1) audience identification, (2) key message extraction, (3) outline co-creation, (4) slide generation with layout suggestions, (5) avatar narrates each slide for review, (6) conversational iteration, (7) export to PDF/PPTX. Evaluate Reveal.js vs. Slidev for in-app rendering and pptxgenjs for PPTX export. Create a Phase 2 story in the existing epics.