You built a pricing card in v0. The layout is close, but the CTA button sits too far from the price and looks orphaned. You do not want v0 to regenerate the whole component and undo the three tweaks you already like. You want it to change one thing.
The trick is to give v0 a tight visual reference and a single instruction, in a format it reads cleanly. Here is the order of operations.
Step 1: Open the preview and capture the screen
With your v0 preview open in a browser tab, go to start a new review and click Capture screen. The browser asks which window or tab to share, you pick the preview, and the current frame is drawn to a canvas. Nothing installs. There is no extension to approve and no signup gate in front of you.
The outcome: you have a still of the exact state of the component, including the spacing problem, frozen in place. A live page keeps changing under you; a still does not.
Step 2: Crop down to the subcomponent, not the page
Drag a rectangle around just the pricing card. Leave out the nav, the footer, and the three cards next to it. Cropping the captured still is the one image edit here, and it does real work: it tells v0 which element you mean before you write a single word.
If you hand v0 the whole page, it has to guess which of a dozen regions you care about. A cropped still removes that guesswork. This is the same reason a cropped still beats an annotated live page, and it is why the crop matters more than any label you could add. For a fuller breakdown of what to include, see what to crop in and out of a screenshot for an agent.
The outcome: one screenshot item that shows the pricing card and almost nothing else.
Step 3: Add a pin to the exact spot
Drop a numbered pin on the gap between the price and the button. A pin points at coordinates, so v0 does not have to parse a sentence like "the space under the number near the bottom." The pin says here.
The outcome: your item now carries a screenshot, a crop, and a pin marking the precise location of the change.
Step 4: Dictate the one change you want
Click into the comment field and dictate instead of typing. In Chrome or Edge the browser's built-in speech recognition turns your voice into text, so you can say the change the way you would say it out loud: "Reduce the vertical gap between the price and the Get started button to about 8 pixels so they read as one group."
Keep it to the single change. Resist the urge to also mention the font weight and the border radius you might revisit later. One item, one requested change, so v0's next turn does exactly that and nothing else. If you are on Firefox, dictation is not available, so you type the comment instead. Dictation shines when the instruction is long, which is covered in when dictation beats typing for a bug report.
The outcome: a screenshot with a pinned spot and one plainly worded instruction attached to it.
Step 5: Publish and hand v0 the markdown link
Click Publish. The review saves and gets a short public URL at /r/<slug>. Add /markdown to that URL and you get the same review as plain-text markdown: the pin, the location, and your instruction, in a format an agent reads without choking on layout. That markdown link is what you paste into v0.
The point of the whole v0 change-request flow is that v0 gets a reference image and a scoped instruction, not a vague ask that triggers a rebuild. Paste the markdown, and v0 has the picture, the exact spot, and the one thing to change. It edits the element instead of regenerating the component. This is the same reasoning behind giving v0 a visual change request without a rebuild.
The outcome: v0 makes the single spacing change and leaves the rest of your card alone.
Why the markdown link matters more than the screenshot alone
A screenshot pasted straight into a chat gives v0 pixels and nothing else. The markdown export carries the structure: which item, which pin, what you asked for, in reading order. That is what agent-readable feedback means in practice. If you want to see the export format itself, capturing a screen to markdown walks through it end to end.
The link is not only for v0. Anyone with it can open the same review as a page, add a comment on the item, and you can mark that comment resolved when the change lands. So a teammate reviewing the card sees exactly what you sent to v0, no separate write-up needed.
When you have more than one change
Add a second item. Capture again, crop the next subcomponent, pin it, dictate its instruction. Each requested change becomes its own item in the same review, and the markdown lists them in order. v0 works through them one at a time instead of trying to interpret a paragraph that packs five asks into one breath.
The whole thing is free, and you can publish without an account. If you sign in later with Google or a magic link, an anonymous review you already published gets claimed to your account, as long as you do it inside the 30-day window. Start your first one at the v0 workflow page and send the change instead of describing it.