Narrate Β· Illustrate Β· Present
SlideStream turns a Markdown or PowerPoint deck into a narrated video presentation β written as real spoken narration (not read-aloud bullets), voiced by a clone of your own voice, illustrated with AI images, and optionally fronted by a talking-head avatar of you.
Open source. Self-host it, run it on your laptop, or bring your own AI keys β your decks and your voice stay yours.
A Markdown file or a PowerPoint .pptx β speaker notes become the narration source when present.
A 20-second voice sample and a photo (both optional). Used once per render, never stored.
Presenter-style narration, AI or stock images per slide, and an optional talking head β rendered to MP4.
Self-hosted Chatterbox voice cloning from a short sample, offline Kokoro, or bring your own ElevenLabs / OpenAI key. Never reads the slides aloud β it writes real narration.
Local folder, self-hosted SwarmUI/FLUX, Google Imagen, DALLΒ·E, or free Pexels/Unsplash stock β matched to each slide.
A lip-synced head of you in the corner β self-hosted SadTalker, or hosted D-ID. Optional, off by default.
Presenter-style from notes or content, a length target per slide, verbatim speaker notes, or your own exact script.
Runs on your machine or your server. Voice samples and photos are used per-render and deleted β no biometric data at rest.
enrich a deck with images (Markdown or PPTX), or create the full narrated video.
Free and open source. The desktop app runs everything on your machine β no account, no terminal.
macOS: the build is unsigned for now β first launch via right-click β Open. Prefer the terminal? The pip install is on the right. View source.
# everything, including the web UI
pip install "slide-stream[all-ai,serve]"
# make a video
slide-stream create deck.md out.mp4
# or open the web UI in your browser
slide-stream serve
Python 3.10+ and FFmpeg required. Provider keys are optional and read from your environment.