🎓 Auditorium

The presentation framework for people who think in code.

Auditorium lets you build live technical presentations as Python scripts. Each slide is an async def function. Animate algorithms step by step, render live-computed plots, run numerical demos — anything Python can do, your slides can do. No PowerPoint. No Markdown. Just code.

from auditorium import Deck

deck = Deck(title="My Talk")

@deck.slide
async def sorting_demo(ctx):
    """Explain how the algorithm builds the sorted prefix."""
    await ctx.md("## Bubble Sort, Step by Step")
    data = [5, 3, 8, 1, 2]
    for i in range(len(data)):
        for j in range(len(data) - 1 - i):
            if data[j] > data[j + 1]:
                data[j], data[j + 1] = data[j + 1], data[j]
            await ctx.md(f"`{data}`")
            await ctx.sleep(0.5)
    await ctx.step()
    await ctx.md("**Sorted!**")

pip install auditorium
auditorium run talk.py

✨ Why Auditorium?

Most presentation tools treat slides as static documents. Auditorium treats them as programs.

• 🐍 Run algorithms live — sort arrays, traverse graphs, train models, all animated on stage
• 📊 Compute content — generate plots, tables, or LaTeX from data, not screenshots
• 📦 Use any Python library — numpy, matplotlib, pandas, whatever you import works
• 🌍 Share with students worldwide — presenter mode syncs your slides to every connected browser in real time
• 🔗 Go public — --public gives you an instant shareable URL, no deployment needed
• 📤 Export everywhere — record to video, export to PDF, or share as a self-contained HTML that replays your talk with original timing

If you've ever wished you could await inside a PowerPoint slide, this is for you.

🚀 Quick Start

pip install auditorium    # or: uv add auditorium

Create talk.py:

from auditorium import Deck

deck = Deck(title="My Talk")

@deck.slide
async def intro(ctx):
    """Notes for the presenter — only visible in presenter view."""
    await ctx.md("# Welcome!")
    await ctx.md("*Press right arrow to continue*")

@deck.slide
async def demo(ctx):
    """Show progressive reveals and timed content."""
    await ctx.md("## Key Points")
    await ctx.step()
    await ctx.md("- First point")
    await ctx.step()
    await ctx.md("- Second point")
    await ctx.sleep(1)
    await ctx.md("*(that one appeared automatically)*")

Run it:

auditorium run talk.py

📋 Features

🌍 Share Publicly

Present from your laptop, share with the world:

auditorium run talk.py --public

╭───────────────────────── Auditorium ─────────────────────────╮
│ Deck:   My Talk                                               │
│ Slides: 15                                                    │
│ URL:    http://127.0.0.1:8000                                 │
│ Mode:   independent (per-tab)                                 │
╰──────────────────────────────────────────────────────────────╯

Public URL: http://vps.apiad.net:4243/r/my-talk/

Anyone with the link sees your presentation in real time. No deployment, no hosting — your laptop runs the deck, a lightweight relay forwards it.

# Choose your own URL slug
auditorium run talk.py --public --name my-talk

# Use your own relay server
auditorium run talk.py --public --relay myserver.com:4243

Self-host a relay (it's one command):

auditorium relay                    # run directly
make relay-install                  # install as systemd service
make relay-update                   # pull + sync + restart

🎤 Presenter Mode

Start with --presenter to sync all audience tabs to your navigation:

auditorium run talk.py --presenter

Two tabs open: your presenter view (notes + timer + slide mirror + next-slide preview) and the audience view. Navigate from the presenter tab — every connected browser follows in real time.

• 📝 Docstrings become speaker notes (never shown to the audience)
• ⚡ Late-joining tabs catch up instantly (full slide state replayed)
• 🔒 Audience keyboards are locked — only the presenter navigates

Without --presenter, each tab navigates independently.

📐 Layouts

@deck.slide
async def layout_demo(ctx):
    """Layouts nest freely."""
    await ctx.md("## Two Columns")
    left, right = await ctx.columns([2, 1])

    async with left:
        await ctx.md("Main content (2/3 width)")

    async with right:
        await ctx.md("Sidebar (1/3)")

Use "auto" for natural-size regions:

header, body, footer = await ctx.rows(["auto", 1, "auto"])

🧪 Show Any Jupyter Object

ctx.show(...) speaks the Jupyter display protocol. Pass any object that implements _repr_html_, _repr_svg_, _repr_png_, or _repr_jpeg_ and it just renders — no adapters, no bundling, no glue code.

import pandas as pd
import matplotlib.pyplot as plt
from tesserax import Canvas, Circle, Square
from tesserax.layout import RowLayout

@deck.slide
async def live_data(ctx):
    # A pandas DataFrame — _repr_html_
    df = pd.DataFrame({"x": [1, 2, 3], "y": [4, 5, 6]})
    await ctx.show(df)

    # A matplotlib figure — _repr_png_ (or _repr_svg_ with the svg backend)
    fig, ax = plt.subplots()
    ax.plot([1, 2, 3], [4, 5, 6])
    await ctx.show(fig)

    # A tesserax Canvas — _repr_svg_
    with Canvas() as canvas:
        with RowLayout():
            Square(30, fill="green")
            Circle(20, fill="red")
    await ctx.show(canvas.fit(padding=10))

This works with matplotlib figures, pandas DataFrames, altair charts, plotly figures, tesserax canvases, sympy expressions, IPython rich objects, and anything else in the Jupyter ecosystem. Plain strings are still passed through as HTML. Runnable example: examples/tesserax_demo.py.

🎬 Recording & Export

Requires pip install auditorium[record] and playwright install chromium.

# Record to video
auditorium record talk.py -o talk.webm

# Export to PDF (vector), HTML (interactive), or PNG
auditorium export talk.py -f pdf -o talk.pdf
auditorium export talk.py -f html -o talk.html
auditorium export talk.py -f png -o slides/

# Step-by-step: one frame per reveal, with original timing in HTML
auditorium export talk.py -f html --step-by-step -o talk.html

⌨️ Navigation

📚 Example

See examples/demo_deck.py for a full 11-slide deck exercising every feature.

auditorium run examples/demo_deck.py

📜 License

MIT