Editable PPTX Skill
An agent skill for Claude Code, Codex and other SKILL.md-compatible agents that turns
image-based slides into an editable PowerPoint (.pptx): real text boxes, shapes and
pictures on separate layers, instead of one flat picture per slide.
Typical inputs:
- a slide deck exported as an image-only PDF (for example a NotebookLM / Gemini Notebook slide deck);
- screenshots of slides;
- slides generated by image models (ChatGPT / GPT Image, Nano Banana / Gemini);
- any ordered set of PNG, JPG or WebP slide images (one image becomes one slide).
The skill calls the Reslide REST API by Tosea.
New accounts get 10 free pages; after that, 1 page = 1 credit.
| Input: flat slide image | Output: editable PPTX opened in PowerPoint |
|---|
| |
Install
Claude Code: plugin marketplace
/plugin marketplace add ToseaAI/editable-pptx-skill
/plugin install editable-pptx@tosea-skills
Claude Code: copy the folder
git clone https://github.com/ToseaAI/editable-pptx-skill.git
mkdir -p ~/.claude/skills
cp -r editable-pptx-skill/skills/pptx-convert ~/.claude/skills/
Use .claude/skills/ inside a project instead of ~/.claude/skills/ to scope it to that project.
Codex and other agents
If your agent reads SKILL.md skills, copy skills/pptx-convert into its skills directory
(for Codex CLI: ~/.codex/skills/). Otherwise, add the contents of
skills/pptx-convert/SKILL.md to the agent's instructions and keep
scripts/convert.py somewhere the agent can run it.
Set up an API key
-
Sign in at https://reslide.tosea.ai and create a key at https://reslide.tosea.ai/keys.
-
Export it in the shell your agent runs in:
export TOSEA_PPTX_API_KEY=tpx_live_...
Do not paste the key into the chat. The skill tells the agent never to print, log or commit it.
Use it
Ask in plain language:
- "Make
notebooklm-deck.pdf an editable PowerPoint."
- "These six screenshots are my slides. Turn them into one editable pptx, in this order."
- "Turn these Nano Banana slide images into an editable deck called
q3-review.pptx."
Or run the script directly (Python 3.8+, standard library only, no pip install):
python skills/pptx-convert/scripts/convert.py deck.pdf --out deck.pptx
python skills/pptx-convert/scripts/convert.py slide1.png slide2.png slide3.png
python skills/pptx-convert/scripts/convert.py --pdf-url https://example.com/deck.pdf
The script uploads the files, creates the task, polls until it finishes and saves the .pptx.
The saved path is the only line printed to stdout. Re-running the same job reuses the existing task
instead of charging twice. See SKILL.md for every option and error code.
Limits and billing
Checked against the service on 2026-11-02; query GET /v1/limits for the current values.
- Inputs: PDF, PNG, JPG/JPEG, WebP. Default limits: 100 MB per upload, 100 pages per task.
- 1 credit per selected PDF page or source image. Credits are reserved when the task starts and
settle only on success; failed or canceled tasks cost nothing.
- Free: 10 pages, one time. Paid plans start at $19/month for 300 pages
(pricing).
- The download link for a finished deck is valid for 24 hours.
What to expect
This is a reconstruction from pixels. Text, shapes and images become editable, but anything that is
not visible in the input (speaker notes, animations, links, chart data, original fonts) cannot be
recovered. Check small text, numbers, charts, tables and formulas before you send the deck on.
No API key? Use the web tools
Related
License
MIT
