You have a 42-minute recording of a usability session with a participant named Dana. She struggled to find the export button, misread the pricing table, and gave up on the account settings page. Right now that struggle lives in your head and a scattered text file. Your PM wants the takeaways by Thursday, and two of the fixes need to reach a developer.
Here is how to turn that raw session into a single review link, using nothing but a browser tab. No install, no extension, no signup required to publish.
Set up: the recording on screen, Cobalt Capture in a tab
Open your video file (or the recorded call) in one window. In another tab, go to start a new review. Play the recording to the first moment that matters. For Dana that is 4:12, where she hovers over the toolbar for eleven seconds looking for export.
Pause the video there. Click Capture screen, and the browser prompts you to share the window with the recording. The current frame gets drawn to a canvas. Drag a rectangle around the toolbar so the screenshot is just that region, not the whole desktop. Cropping the still is the only image edit, and that is fine here. You want the toolbar and Dana's cursor, nothing else.
Add the commentary the way you would say it out loud
Each screenshot becomes an item. Instead of typing, click the dictate button and talk through what happened. In Chrome or Edge the browser's speech recognition writes it down as you speak. For the toolbar capture you might say:
Dana spent eleven seconds hovering here looking for export. She read the download icon as import and never clicked it. Two other participants did the same thing. The label needs to say Export, not just an icon.
That is one item done. Add a numbered pin on the download icon so anyone reading knows exactly which control you mean. Then scrub the recording to the next moment, 17:30, where Dana misreads the pricing table. Capture, crop to the table, dictate again:
She thought the middle column was the free plan because the highlight color pulled her eye there. She said, quote, so this one is free, right. It is the forty-nine dollar plan. The visual emphasis is on the wrong tier.
Repeat for the account settings dead end at 31:05. Three moments, three items, each with a cropped still, a pin, and a spoken note. If you would rather see dictation broken down on its own, there is a full walkthrough of dictating your review comments.
Add findings that have no screenshot
Not every insight points at a pixel. Dana said twice that she expected a confirmation email and never got one. There is no screen to capture for that. Add a free-floating comment item with just the text: "Participant expected a signup confirmation email; none arrived. Check the transactional email flow." It sits in the same review alongside the visual items.
Publish: one link the whole team reads
Click Publish. The review saves and gets a short public URL like /r/dana-session-1. Anyone with that link opens it in a browser and reads your three visual findings plus the email note, in order, with the cropped stills and pins intact. No login on their end.
The same review is also available as a plain-text markdown version at /r/<slug>/markdown, and you can export it as a PDF or a Word document. So the researcher who wants a tidy PDF for the readout deck gets one, and the developer who wants to paste the export label finding into their editor gets clean text. This split output is the core of turning sessions into a research synthesis your team can act on without three different tools.
What your teammates do with the link
Anyone with the review link can comment on any individual item. When your PM reads the pricing finding and asks whether it repeated across sessions, they reply right on that item. You can mark each comment resolved as you address it, so the thread stays honest without you standing up a bug tracker. If you want the mechanics, see how to resolve comments without a bug tracker.
You also get an analytics page on your side showing visit counts and rough visitor geography, so you know the design lead actually opened it before the readout.
Doing this across five sessions
One interview gives you one review. Five interviews with the same script give you five reviews, and the patterns jump out fast: if four of five participants misread the pricing highlight, that is not a fluke, that is a fix. Keep the item wording consistent across reviews ("pricing highlight, wrong tier") so you can scan them side by side.
A few things to keep the captures clean. Pause the recording before you capture so the frame is sharp. Crop tight to the control in question. Say the timestamp in your dictation so anyone can jump back to the source clip. And keep one finding per item, not three crammed together, so each note stays actionable. More of these habits are collected in the notes on what makes screen feedback hard to act on.
If your reviewer or a teammate is on Firefox, dictation is not available there and they type instead; everything else works the same. Open a tab, play the recording, and capture the first moment that mattered. By the time you reach the end of the tape you have a link ready to send.