Image to PowerPoint — screenshot → editable .pptx
Updated 2026-04-20 · ~6 min read · Tool + guide
slideforge.dev turns a screenshot, whiteboard photo, or reference image into a real editable PowerPoint slide — not a pasted picture. A vision-capable LLM identifies text, shapes, and charts; our renderer outputs native .pptx shapes you can edit, translate, or rebrand. ~30 seconds per slide, $0.20, first ~15 conversions free on signup.
Drag a PNG, JPG, or WebP image here, or
Max 20 MB. Screenshots of existing slides, competitor decks, whiteboards, sketches, or reference imagery all work.
TL;DR
- Screenshots → editable .pptx, not pasted images. Text becomes text boxes, charts become editable charts.
- Works on competitor slides, whiteboard photos, sketches, Figma exports, reference decks.
- Cost: $0.20 per generated slide, ~30s. $3 free credit covers ~15 conversions.
- Available via the playground, REST API, and MCP server (
create_slide(from_image=...)).
What it's for
Rebranding a client's existing deck
Client sends screenshots of their legacy deck and asks for an updated version in the new brand. Instead of recreating each slide by hand, convert the screenshots to editable .pptx, apply your brand theme, and the typography + colors snap into place.
Translating a deck where only screenshots exist
You have screenshots of a Japanese deck and need an English version. Convert via image-to-pptx first (so every text element becomes a real text box), then pass through translation. Fonts and layouts survive both steps.
Whiteboard → polished slide
Photograph a workshop whiteboard with boxes, arrows, and handwritten labels. The vision LLM identifies the diagram structure and recreates it with clean shapes, consistent typography, and grid-aligned positions. Hand-drawn becomes presentable.
AI agent that analyzes and redraws decks
An agent working in Claude Desktop or Cursor receives a customer's deck as screenshots, identifies key slides, and recreates them in the customer's brand. Call create_slide(from_image={image, instructions}) via MCP — the agent gets back an editable .pptx URL.
How image-to-PowerPoint works
- Upload an image via the drop-zone above (PNG, JPG, WebP; max 20 MB). Higher resolution = better text extraction.
- Vision LLM analyzes the layout. The model identifies text regions, shapes, charts, and their positions.
- Spec generation. The model emits a JSON component spec — e.g.
{component: 'split_view', left: {component: 'metric', ...}, right: {component: 'bar_list', ...}}. - Native .pptx render. Our spec renderer produces a real PowerPoint file with editable shapes, text boxes, and charts.
- Apply your theme (optional). Upload a corporate .pptx template; your fonts and colors replace the image's defaults.
Image to PowerPoint — tool comparison
Related: PDF to PowerPoint · PPTX Generator · PowerPoint API guide
Convert image to .pptx via API
Call create_slide with from_image — works via REST or MCP.
# REST — base64-encoded image
curl -X POST https://api.slideforge.dev/v1/generate \
-H "Authorization: Bearer sf_live_..." \
-H "Content-Type: application/json" \
-d '{
"from_image": {
"image": "data:image/png;base64,iVBORw0KGgo...",
"instructions": "Recreate in our corporate blue theme"
}
}'
# Returns 202 with job_id. Poll GET /v1/jobs/{id} for the .pptx URL.# MCP (Claude Desktop, Cursor, etc.)
create_slide(
mode="auto",
from_image={
"image": "<base64>",
"instructions": "Recreate with our brand colors"
}
)Frequently asked questions
How does image-to-PowerPoint work?
Upload a screenshot, photo, or reference image. slideforge.dev runs a vision-capable LLM (Pixtral Large / GPT-5.4 depending on the route) to analyze the layout — identifying text, shapes, charts, and their positions — then generates a JSON component spec that our renderer turns into a native PowerPoint (.pptx) file. Output has real editable text boxes and shapes, not rasterized pixels. Cost: $0.20 per generated slide; ~30 seconds end-to-end.
What kinds of images work best?
Any image with clear slide-like structure works — screenshots of existing slides (yours or a competitor's), whiteboard photos, architecture diagrams, sketches on paper, reference images from design tools. Resolution matters: 1280px+ along the longest edge gives better text extraction. Low-contrast or handwriting-only images work via OCR but with less fidelity. For scanned documents, try the PDF-to-PPTX tool instead.
Does it preserve fonts and exact colors?
The engine reconstructs the layout with close-match fonts from PowerPoint's standard font set and extracts dominant colors from the source image. It's a recreation, not a pixel-perfect clone — fonts may differ slightly, spacing and color saturation are approximations. For exact brand adherence, apply your theme via /v1/themes after generation and the colors + fonts snap to your corporate palette.
Can I screenshot a chart and get an editable chart back?
Yes for common chart types — bar charts, line charts, pie/donut, waterfalls, simple funnels, org charts. The vision LLM identifies the chart type and values, then routes to the matching component (BarList, LineTrend, Donut, Waterfall, OrgChart, etc.). You get a native PowerPoint chart you can edit — change values, recolor, resize. For uncommon or custom chart types, the output is a shape-based approximation.
What does image-to-PowerPoint cost?
$0.20 per generated slide. The $3 free credit on signup covers ~15 conversions. No subscription — USD wallet with $10 minimum top-up and volume discounts up to +20% at $200+ deposits.
Is there an image-to-PowerPoint API?
Yes. POST the image (as base64 or a URL) to https://api.slideforge.dev/v1/generate with the from_image parameter, or call the create_slide MCP tool with from_image={image: base64, instructions}. Both route through the same vision-first pipeline. See slideforge.dev/docs/api for the full reference and slideforge.dev/docs/mcp for agent integration.
How does this compare to just pasting an image into PowerPoint manually?
Pasting drops the image as a picture object — the text inside the image isn't editable, you can't change colors, you can't repurpose individual shapes. slideforge.dev's conversion extracts each element as a native PowerPoint shape: text is text, chart data is data, boxes are boxes. Essential if you need to rebrand, translate, or edit the recreated slide.
Can an AI agent use this?
Yes — this is one of the most common patterns in our logs. An agent analyzes a customer's existing deck (screenshots uploaded via chat), extracts key slides, and recreates them in the customer's brand template. All via the create_slide MCP tool with from_image — no human in the loop required. See the MCP docs for the full workflow.
Recreate your first slide from an image in under a minute.
$3 free credit covers ~15 conversions. No subscription. No credit card.
Canonical: slideforge.dev/tools/image-to-pptx. Published by Smart Data Brokers GmbH, Zurich. Not affiliated with slideforge.io, slideforge.fr, or unrelated GitHub repositories of the same name.