My AI Text to Slide Generator

September 06, 2021 09:12 AM Share

This project turns free-form text into a branded slide deck using Streamlit, OpenAI, or an offline rule-based generator, and converts the result into a downloadable PowerPoint using python-pptx.

What this app does

Accepts a topic/paragraph and a brand color.
Generates a deck JSON (with titles and bullets) using either OpenAI or an offline rule-based generator.
Scores the result for basic quality.
Converts the JSON into a downloadable PowerPoint (.pptx).

File 1 — app.py

Imports and configuration

The top imports bring in streamlit, typing helpers, json, time, and utility modules. The app later calls into utils_pptx.py to build the PowerPoint.

1) Call-rate limiting

COOLDOWN_SECONDS = 8
DAILY_LIMIT = 60
HOURLY_SHARED_CAP = 200

def init_rate_limit_state():
    ...

def can_call_now() -> Tuple[bool, str, int]:
    ...

def record_successful_call():
    ...

def _shared_hourly_counters():
    ...

def _hour_bucket(ts=None):
    ...

Ensures API usage is controlled and user-friendly.

2) Content shape constraints

MAX_TITLE_WORDS = 12
MIN_CONTENT_SLIDES = 4
MAX_CONTENT_SLIDES = 10
MIN_BULLETS = 3
MAX_BULLETS = 6
MAX_BULLET_WORDS = 16

Defines structural rules for slide content.

3) Data models

class SlideModel(TypedDict):
    title: str
    bullets: List[str]

class DeckJSON(TypedDict):
    title: str
    subtitle: str
    slides: List[SlideModel]

class Slide(NamedTuple):
    title: str
    bullets: List[str]

class Deck(NamedTuple):
    title: str
    subtitle: str
    slides: List[Slide]

Schema for JSON interchange and internal deck representation.

4) OpenAI generation

def call_openai(topic: str, model: str, temperature: float, max_tokens: int) -> str:
    # Crafts system prompt and calls OpenAI API
    return response_text

Generates a JSON deck from OpenAI models.

5) Text utilities

def naive_sentence_split(text: str) -> List[str]:
    ...
def extract_keywords(text: str, k: int = 6) -> List[str]:
    ...
def chunk_list(xs: List[str], n: int) -> List[List[str]]:
    ...

Support functions for offline deck generation.

6) Offline deck generator

def generate_offline_deck_json(topic: str, brand_hex: str) -> DeckJSON:
    ...

Produces a structured deck without external APIs.

7) JSON hardening

def coerce_json_block(s: str) -> str:
    ...

Extracts and sanitizes valid JSON from model outputs.

8) Deck scoring

def score_deck(dj: DeckJSON) -> Tuple[float, dict]:
    ...

Computes quality heuristics for slide content.

9) Streamlit UI

st.set_page_config(page_title="AI Text → Slides", page_icon="📑", layout="wide")

provider = st.selectbox("Provider", ["OpenAI", "Offline (rule-based)"])
model = st.selectbox("Model (OpenAI)", ["gpt-4o-mini", "gpt-4o", "gpt-4.1-mini"])
brand = st.color_picker("Brand color", value="#0F62FE")
temp = st.slider("Creativity (OpenAI)", 0.0, 1.0, 0.4, 0.05)
max_tokens = st.slider("Max tokens (OpenAI)", 256, 2048, 900, 32)

topic = st.text_area("Enter topic or paragraph", height=160)

if st.button("Generate Slides"):
    ...

Controls UI, model selection, and output display.

File 2 — utils_pptx.py

1) Color helpers

def hex_to_rgb(hex_color: str) -> Tuple[int, int, int]:
    ...
def apply_brand_to_title(shape, rgb):
    ...

Handles brand coloring.

2) Slide builders

def add_title_slide(prs, title, subtitle, rgb):
    ...
def add_content_slide(prs, title, bullets):
    ...

Adds title and content slides.

3) Deck to PPTX

def deck_to_pptx_bytes(deck: Deck, brand_hex: str) -> bytes:
    ...

Assembles and exports deck to .pptx bytes.

End-to-end flow

User enters text and color.
Deck is generated (OpenAI or offline).
JSON is validated and scored.
Deck is converted into .pptx.
User downloads slides.

Run locally

pip install -r requirements.txt
streamlit run app.py

For OpenAI mode:

export OPENAI_API_KEY="sk-..."

Offline mode works without internet.