Vwrap's tools: research a niche, generate the thumbnail, edit one layer of it.

How the three stages connect

Two hand-offs join them, and a person triggers both.

The first is Create Similar, a button on every outlier card and on each video in a channel profile. Pressing it opens the create page with that video's thumbnail already attached as your reference image and its title carried along with the job, so the pattern you just found is the starting point instead of a note you retype.

The second is the Template slot. Templates are layout archetypes mined from outlier thumbnails in a sub-niche, and one applies only when you pick it. Nothing picks a template for you.

That is the whole of the connection. The generation pipeline receives a selfie, a reference, a template, props, a prompt, a text toggle, an output count, and the source video's title when you arrived from Create Similar. It never queries our outlier rankings, so a thumbnail is shaped by what you chose to send it and not by whatever happened to be trending in your niche that week.

Which one you need

Start from the question you arrived with.

If you do not know what to make yet, start at outlier research. It answers what already over-performed in a niche, measured against each channel's own baseline rather than a platform average, so a small channel's breakout is not buried under a large channel's ordinary upload.

If you know the shot and need the image, go to the thumbnail generator. If the shot is you, the selfie page covers the same generator with the Selfie slot filled, and explains what happens to the photo you upload.

If you already have a thumbnail from a Vwrap run and one part of it is wrong, the editor changes that part by name.

Niche pattern extraction is the odd one out. It is not a screen you open. It documents the research pass that produced the template library, so you can judge how much a template is worth before you trust one.

What one finished thumbnail costs

Research spends nothing. Searching a niche, opening a card, reading a baseline and browsing templates are all free; credits pay for images.

Generation bills per output rather than per run. On a 20-credit model, a run of four outputs costs 80 credits, not 20. The five models run from 15 to 65 credits per output, and the dearest of them is selectable on Pro and Studio only.

Editing is flat. Most operations cost 15 credits and an upscale 10. Retyping the headline costs nothing, because the headline comes back as an editable text layer rather than a re-render.

A worked example: two single-output attempts on a 20-credit model plus one edit is 55 credits. That is the figure the plan sizes are set against, which puts Starter at roughly 45 finished thumbnails a month, Pro at 100 and Studio at 273.

Access, caps and credits

What the toolkit does not do

• The editor opens thumbnails generated in Vwrap. There is no route that imports an outside image as the thing being edited. An outside thumbnail enters through the create page's Reference slot instead, and that produces a new generation rather than an edit of the original.
• Every editor operation re-renders the whole frame. The model is asked to leave the rest alone and nothing enforces it, and the result is re-encoded on upload, so untouched pixels are not preserved byte for byte.
• Layer detection is best-effort. It lists at most 8 layers, drops very small boxes and merges overlapping ones, so a busy frame can leave a prop unlisted.
• We do not have your click-through rate. CTR belongs to the channel owner and is in no public API, so every figure on these pages is a view-based measure.
• The free plan after the trial runs 2 niche searches a day and returns the top 4 results of each. Paid plans raise that daily cap rather than removing it: 10 searches a day on Starter, 30 on Pro, 60 on Studio.
• Long-form and Shorts baselines are never mixed. A channel that posts both carries two separate baselines, and a video is only ever compared against its own format.
• Multi-layer prompt edits are unsupported. One layer plus a prompt works, and so does a free-form prompt over the whole image, but naming two layers in a single edit does not.