Presentia

Decks from documents, data and markdown.

Presentia is the deck satellite of the Escriba family: an open-source, self-hosted AI presentation generator. Turn a prompt, a document, a dataset or a markdown file into an editable presentation and export it to PPTX or PDF. On top of its upstream it adds charts that can’t hallucinate (every number is validated against your data), a per-deck LLM cost panel, an optional PII-anonymization gateway, a guided model picker and a UI in 7 languages.

Presentia is a fork — and says so

Presentia doesn’t reinvent the generator: it is a fork of Presenton, an excellent open-source AI presentation generator. All the core generation machinery is theirs; the fork adds the Escriba Suite integrations, the grounded-charts guard, the cost panel, the model picker, the markdown editor, the i18n layer and the suite branding. See License.

Who it is for

• Anyone who turns notes, documents or reports into slides and wants the AI to run on their own server, exporting real, editable PPTX and PDF files.
• Analysts and accountants who need charts they can defend — every figure in a data-driven deck must exist in the dataset or be an exact column aggregate.
• Markdown writers who want Gamma-style deck building: paste a .md file and choose whether the AI preserves your text verbatim, condenses it or rewrites it.
• Self-hosters and teams who want to pick their own model (including local Ollama), see exactly what each deck cost, and optionally anonymize PII before anything reaches a provider.

Key features

Prompt to deck

Describe the topic, review the generated outline, pick a template and get a full presentation — layouts, text and images — ready to fine-tune in a complete editor with an AI assistant.

Documents to deck

Upload PDF, Word, PowerPoint, spreadsheets, images or plain text and build the deck from their content. Optionally delegate parsing to Escriba for best-in-class conversion with OCR.

Markdown to deck

Every section split by --- or by headings becomes a card. Three text modes — Preserve (verbatim), Condense, Generate — plus a built-in markdown editor with live per-card preview and drag & drop.

Charts that can't hallucinate

Generate a deck from a CSV/TSV/JSON dataset. Every chart figure is validated against the data — exact values or exact column aggregates only. Wrong numbers get one retry with feedback; if the model insists, the slide is rejected.

LLM cost panel

Every model call is logged and attributed to its presentation, stage and slide, priced with a versioned catalog. See what each deck actually cost and compare providers.

Guided model picker

Models ranked by curated quality and blended price, with recommendation badges. One OpenRouter key unlocks almost the entire catalog; switching models is one click.

PII gateway (optional)

With Anonimal enabled, user content and extracted document text are anonymized before reaching the LLM provider. Fail-closed: if the anonymizer is down, generation stops instead of leaking raw PII.

Web search grounding

Ground generations with Searchgirl (the suite’s private metasearch), SearXNG, Tavily, Exa or Brave — or use the model’s native grounding.

Bring your own model

OpenAI, Anthropic, Google, DeepSeek, OpenRouter, Ollama / LM Studio (local), LiteLLM, Azure OpenAI, AWS Bedrock, Google Vertex or any OpenAI-compatible endpoint.

Templates, themes and 7 languages

Built-in template families, custom templates generated from your own PPTX, a full theme editor — and the whole UI in English, Español, Français, Português, Italiano, 中文 and 日本語.

Get started

Install Presentia View on GitHub