Tech Products intermediate 3 min read May 22, 2026 · Updated May 23, 2026
Public Preview Sign in free for the full digest →

Presenton: Self-hosted AI deck generator with editable PPTX export

“5,717 stars and 56 releases in, every single one is still tagged -beta — but the LLM provider list now spans OpenAI, Anthropic, Google, Vertex, Azure, Ollama, and any OpenAI-compatible endpoint.”

Presenton: Self-hosted AI deck generator with editable PPTX export
3 Views
0 Likes
0 Bookmarks
Source · github.com

“"a genuinely solid option ... more capable than I expected for something I spun up in Docker" — XDA Developers review (https://www.xda-developers.com/self-hosted-free-presentation-tool-does-everything-canva-does/, verified 2026-05-22)”

You know that feeling when a Monday morning ask for a 10-slide deck eats your afternoon, and the AI tools that promise to fix this either lock your content into a SaaS or refuse to export clean PPTX? Gamma now ships at $100M ARR partly because that pain is universal, but for anyone in regulated or air-gapped work, sending raw content into a cloud deck builder is a non-starter. The existing open-source path is either reveal.js (HTML decks your boss cannot edit in PowerPoint) or python-pptx scripts (no AI, no design). Neither closes the loop from prompt to editable PPTX without uploading your slides to someone else's server.

aiopen-sourcepresentationsself-hosteddockerfastapinextjs

You launch one Docker container that bundles a FastAPI backend and a Next.js frontend, then point it at any LLM you already use (OpenAI, Anthropic, Google, Azure, Ollama, or a custom OpenAI-compatible endpoint). When you POST to /api/v1/ppt/presentation/generate with text, slide count, language, and template name, the LLM fills HTML+Tailwind templates, the backend renders them, and headless LibreOffice converts the result to PPTX or PDF. The clever bit is using web styling instead of python-pptx — designers can author templates in HTML, which is far more expressive than slide-DSL approaches. The trade-off shows up at the export boundary, where the LibreOffice step is the failure point users hit in open issues.

01
Bring-your-own-LLM — swap between OpenAI gpt-4.1, Anthropic claude-3-5-sonnet, Google Gemini, Azure, Vertex, Ollama llama3.2:3b, or any OpenAI-compatible endpoint via one LLM env var. v0.8.5-beta added LiteLLM, widening that surface even f...
02
Editable PPTX output — exports to .pptx (default) or .pdf, so the result opens in PowerPoint, Keynote, or Google Slides instead of being trapped in a web viewer.
03
HTML+Tailwind templates — author your own designs as web pages rather than slide DSLs, including AI-generated templates derived from existing PowerPoint files.
04
HTTP API with Basic auth — single endpoint POST /api/v1/ppt/presentation/generate returns presentation_id, file path, and edit URL, so you can wire deck generation into other tools.
05
Image pipeline with seven providers — DALL-E 3, Gemini Flash, GPT-Image 1.5, Pexels, Pixabay, ComfyUI, or Open WebUI, with DISABLE_IMAGE_GENERATION as an opt-out.
06
Self-hosted with no telemetry by default-on but easy opt-out via DISABLE_ANONYMOUS_TRACKING, plus the homepage promise: your data is not used to train any models.
07
Cross-platform delivery — same backend ships as a Docker image (ghcr.io/presenton/presenton:latest) and as Electron desktop apps for Windows, macOS, and Linux.
Who it’s for

If you are an engineer at a company that cannot send slide content to Gamma's cloud — healthcare, finance, gov, anyone with strict data residency — and you want a one-Docker-command path from prompt to editable PPTX, this is squarely for you. Same story if you want to wrap deck generation behind your own API for an internal tool. Skip it if you need rock-solid production reliability today: every release since 0.7.x is tagged -beta, the LibreOffice conversion step has open bug reports, and the contributor count is 10.

Worth exploring

Worth exploring if you have a real self-hosting requirement or want to build deck generation into your own product via API — the architecture is sensible and the release cadence is high. Treat it as beta software, not as a Gamma replacement for non-technical users: open issues #564, #571, and #544 all sit on the export path, and the entire 0.8.x line is still beta-tagged. A weekend Docker pilot is the right level of commitment right now.

Developer playbook
Tech stack, code snippet, sentiment, alternatives.
PM playbook
Adoption angles, user fit, positioning.
CEO playbook
Traction signals, ROI, build vs buy.
Deep-dive insight
Full long-form analysis, no fluff.
Easy mode
Core idea, fast — when you need the gist.
Pro mode
Technical nuance, edge cases, tradeoffs.
Read the full digest
Go beyond the preview

Deep-dive insight, Easy and Pro modes, plus action playbooks — the full breakdown is one tap away.

Underrated tools. Unfiltered takes.

Read the full digest in the Snaplyze app for deep-dive insight, Easy and Pro modes, and the playbooks you can actually use.

Install Snaplyze →