Narrate · Illustrate · Present
Turn your slides into a video that speaks in your voice.
SlideStream turns a Markdown file, a PowerPoint deck or a document into a narrated video, an editable PowerPoint, or both. It writes what a presenter would say rather than reading the bullets, voices it offline or in a clone of your own voice, and can add pictures and a presenter in the corner.
Open source and free. No account and no keys needed: the voice runs on your machine, and every AI service is optional.
From deck to video in one step
Bring your content
A Markdown file, a PowerPoint .pptx with speaker notes, or a Word or PDF document that SlideStream drafts into a deck.
Choose a voice and presenter
The offline voice works straight away. Add 10 to 30 seconds of your own voice, and a mascot or photo, if you like. Samples are used for one render and deleted.
Get a video or a deck
An MP4 with narration, pictures and a presenter, or an editable PowerPoint with the narration in the speaker notes.
Everything is swappable, and yours
A voice that works offline
Kokoro speaks on your machine with no keys. Clone your own voice on a self-hosted Voicebox or Chatterbox server, or use ElevenLabs or OpenAI.
Narration, not read-aloud
An LLM of your choice (local Ollama or a cloud model) writes what a presenter would say, fitted to a length. Or speak your notes or your own script exactly.
Pictures that fit
Clean text slides, your own picture folder, free stock photos, or AI images from a local SwarmUI server, Google Imagen or DALL·E.
A presenter in the corner
Six built-in mascots or your photo: held still, a cartoon mouth that moves with the voice, or full lip-sync on your own GPU with Wan2.2-S2V.
Code and tables done right
Code blocks keep their indentation with syntax colouring, and tables keep their rows, in the video and the deck.
Designed slides
Keep your own PowerPoint's look in the video, or have a self-hosted Presenton design the deck, with a check that the design kept your content (experimental).
Two outputs, one deck
A narrated MP4, or an editable PowerPoint with AI speaker notes to present live. Both come from the same narration.
Check before you render
slide-stream doctor estimates length, cost and render time, and warns about problems before a slow render starts.
Private by design
Runs on your machine or your own servers. Voice samples and photos are used per render and deleted, with no biometric data kept.
Get it
Free and open source. The desktop app runs everything on your machine, with no account and no terminal.
The desktop builds are not yet signed: on macOS open the app the first time with right-click → Open; on Windows choose "More info", then "Run anyway". All releases.
# everything: AI providers, offline voice, web UI
pip install "slide-stream[all]"
# make a video, or an editable deck
slide-stream create deck.md talk.mp4
slide-stream create deck.md deck.pptx
# or open the web UI in your browser
slide-stream servePython 3.10 or newer. FFmpeg is included. AI provider keys are optional. PyPI · Documentation