“The image is wrong after I click save...”
Open source by Polyform
AI is bad at video.
Give it the parts that matter.
Record a screen walkthrough the natural way. JesSee pulls out the key screenshots and your spoken explanation, then turns them into an editable PDF that AI can understand, debug from, and act on.
Built for Chrome and Safari · Editable local history · Bring your own OpenAI key
▶18:42
JesSee
VISUAL PLAYBOOKPDFUpdate the product image
A clear outcome, the exact steps, and the visual evidence that proves each one.
01
Open image settings
✓
02
Choose the replacement
✓
03
Confirm the new state
✓
01 The communication gap
Video is natural for people.
It is a terrible input for AI.
A walkthrough contains the full story, but most AI tools struggle to process it accurately or efficiently. Uploading every frame consumes enormous context. Uploading a few screenshots loses the sequence and the explanation. Rewriting the whole thing as a prompt means doing the work twice.
Video overwhelms the context
Every pause, repeated frame, and detour competes with the moments the AI actually needs.
Screenshots lose the story
A folder of images cannot explain their order, the action between them, or why each state matters.
Text loses the screen
A long prompt can describe the product, but the AI still cannot see the exact state you are pointing at.
The idea
A useful video is a sequence of important images plus the explanation around them. What if we kept those—and removed everything the AI does not need?
02 From walkthrough to useful context
Record the video.
Give AI the useful version.
Your recording is raw material, not the final input. JesSee connects what you said to what was on screen, breaks it into clear steps, and lets you choose the screenshots that best communicate the story before creating the document.
STEP 01
Show and explain
Choose a tab, window, or screen. Talk naturally while you work. The glowing cursor makes actions clear, and simple shortcuts let you highlight or redact details.
- Narration stays aligned with the screen
- Important clicks become evidence moments
- Private Mode can keep screenshot pixels local
STEP 02
Edit the draft, not a form
JesSee creates the first version inside one visual editor. Rewrite any sentence, turn ideas into lists or callouts, decide whether each source URL should appear, and click a large image to step through every captured alternative.
- Format text as paragraphs, bullets, numbered lists, and callouts
- Keep the source URL on every step—or hide it before sharing
- Step backward or forward through image choices
STEP 03
Return to any walkthrough
Your explanation does not disappear after one download. The local library keeps each recording, transcript, visual story, and selected evidence together so you can revise or export it again later.
- Search every retained walkthrough
- Play the original recording or read its transcript
- Reopen the editable story and download a fresh PDF
STEP 04
Give AI context it can use
Generate one structured PDF with the sequence, explanation, source pages, and visual proof together. Attach it to an AI conversation, product spec, or bug report without uploading a long video or rebuilding the story by hand.
PDF Download example PDFOne continuous playbook made by JesSee from this site ↓
03 Try the developer preview
Open source · v0.1.0 alpha 4
Choose your browser.
See if JesSee helps.
These are early, unverified builds from the public GitHub repository. They are not yet listed in the Chrome Web Store or Mac App Store, so installation takes a few deliberate steps.
DEVELOPER PREVIEW
LOAD UNPACKEDChrome
Download and unzip the extension, then load its folder from Chrome's Extensions page.
- Open chrome://extensions
- Turn on Developer mode
- Choose Load unpacked and select the unzipped folder
TEMPORARY BETA
SAFARI 17+Safari
Safari can load the open-source extension ZIP temporarily while the signed and notarized release is being prepared.
- Enable Show features for web developers
- In Safari Settings → Developer, allow unsigned extensions
- Choose Add Temporary Extension and select the ZIP
Built-in update checks · automatic installation follows store and signing approvalRead the complete installation guide ↗
04 Built for AI, useful beyond it
Explain it once.
Use the story anywhere.
I built JesSee to give AI better product context. The same visual document also makes tutorials, playbooks, and offline guidance easier for people to use.
Specs for AI agents
I walk through the current product, explain the behavior I want, and give the resulting visual spec to an agent. It gets the flow and the evidence—not a folder of unlabeled images.
Debugging and bug reports
I reproduce the issue once, then give an AI or teammate the exact path, the unexpected state, and the screenshots needed to diagnose and fix it.
Tutorials and playbooks
A natural walkthrough becomes step-by-step guidance with the right images already connected to the right explanation.
Thoughtful offline handoffs
I can explain a complex idea at full length, then send a clean document someone can scan and reference instead of asking them to watch an 18-minute recording.
“
I built JesSee because video is often the easiest way to explain a product and one of the worst formats to give an AI. The output should not be more media to process. It should be the screenshots, sequence, and explanation the AI actually needs.
AE
Ahmed ElsamadisiFounder, Polyform
Why I built it
The recording is not the product.
Usable context is.
JesSee keeps the natural act of showing and explaining, then compresses the walkthrough into the moments that carry meaning. The result is durable context that can move cleanly between an AI agent, a teammate, a tutorial, and a product workflow.
Open source · MIT licensed
This is just the start.
Please make it better.
JesSee is an experiment in a better way to give visual product context to AI—and a useful way to turn walkthroughs into tutorials, playbooks, and guidance for people. Try it on real work, report what breaks, or help build what should come next.
How data moves: Recordings, screenshots, and PDFs remain stored on your computer. Creating a plan sends narration and selected visual evidence to OpenAI through the API key you provide; Private Mode omits screenshot pixels. This website uses Google Analytics to measure page visits, downloads, and GitHub-link activity, but never sends recordings, screenshots, narration, API keys, email addresses, or unapproved URL parameters. Only standard campaign-attribution fields are retained, and values that resemble email addresses are discarded.
polyform-ai / jesseePublic
Help AI see what you see.
▣ safari◇ src▤ docs◆ website