Deckrun is not a presentation tool. It is a document execution system.

It takes structured markdown and produces polished PDFs, narrated videos, and audio assets — deterministically, from an HTTP API, for AI agents and backend services.

Inputs and outputs¶

All outputs derive from the same markdown source and theme. One source, reproducible every time.

Core capabilities¶

High-fidelity PDF generation¶

Markdown is parsed into a structured intermediate representation and rendered via a themeable layout engine. Supported layouts: title, two-column, image-heavy, section header, full-blank, and more. Brand-safe colors, typography, and logo placement are controlled by the theme profile schema.

Outputs include: single-column resumes, two-column resumes, pitch decks (internal vs external), datasheets, and whitepapers.

Video output (slides → MP4)¶

Any Deckrun deck can be rendered as an MP4 video. Slide timing is driven by notes duration (TTS) or explicit timer values. The video is rendered from the same deterministic source as the PDF — no visual drift, no narration mismatches.

Narration pipeline¶

Slide notes → TTS synthesis → per-slide audio segments → concatenated MP3 with chapter markers and a CUE sheet. The narration pipeline supports:

• System tier: Professional neutral voices, included with all plans
• Premium tier: High-fidelity expressive or cloned voices (add-on)

The internal TTS backend is not exposed. Deckrun owns timing, segmentation, and muxing. Providers are interchangeable.

Notes text extraction¶

Every slide's presenter notes are captured as clean, reusable text. Usable for TTS, blog posts, summaries, FAQs, or agent analysis.

Async job model¶

Large outputs (video, audio) are generated in the background. Your system receives a job ID and polls GET /jobs/{id} for status and asset URLs. Safe to call from chatbots, workflows, and agent loops without timeouts.

Output connectors¶

Deliver finished artifacts to: S3, Google Drive, Dropbox, Box, internal filesystems, or Slack. Optionally bundle into a single zip (PDF + video + notes + audio). Connectors and bundling are handled by the service.

Agent-friendly schemas¶

Strict, versioned JSON Schema for slides, themes, voices, job status, and outputs. Validate requests before production. Schemas are stable, permanent, and served at canonical URLs.

Example flow¶

1. Agent generates markdown + picks theme
2. POST /generate (markdown, theme, voice_id, output_destination)
3. → 202 Accepted: { job_id: "j_abc123", status: "queued" }
4. GET /jobs/j_abc123 → { status: "success", artifacts: [ pdf_url, mp4_url, mp3_url ] }
5. Artifact URLs delivered to output connector (S3, Drive, etc.)

Determinism and schemas¶

Every Deckrun run is deterministic given the same inputs. Outputs can be reproduced exactly. Schemas are versioned and stable — agents and tools can rely on them without defensive coding against format changes.

See Schemas for the full public schema index.

Async model¶

All heavy outputs (video, audio, large PDFs) use the async job model. Light PDF-only requests may return synchronously. The status field in the job response schema indicates whether the result is synchronous or queued.

What this is not¶

• Not a presentation design tool
• Not an LLM writing system (Deckrun renders; your agents generate)
• Not a general-purpose document converter
• Not a consumer app

Deckrun has one job: given structured markdown and a theme, produce finished artifacts reliably at scale.