Upload your voice guide and the content you want checked. Sprites reads both, marks where tone, vocabulary or claims drift, and rewrites the passages that miss so they pass.
When one lean team runs ads, SEO, email and social, the brand starts to sound like five different people. This template puts a reviewer between drafts and publishing. Sprites learns your voice from whatever you have: a style guide PDF, a few approved pieces, a list of banned words, or a plain description of how you want to sound. Then it reads the content you give it and grades each piece against that baseline.
The check is specific. Instead of a single score, you get line-level findings: this headline is more formal than the guide, this paragraph uses a competitor term you banned, this ad claims something the guide says never to promise, this email switches from "you" to "users" halfway through. Each finding says what was expected, what was found and the smallest edit that fixes it, with a rewrite you can accept or adjust.
The same checker can run on content Sprites produces itself. Ask it to draft ad copy, an article or a newsletter and then check the output against the guide, and the two steps happen in one conversation.
Sprites reads the guide and the approved examples and turns them into concrete rules: tone words, formality range, sentence length patterns, preferred and banned vocabulary, how the brand refers to itself and the customer.
Each piece is read in full, whether it is a 40-character headline or a 2,000-word article. Channel context is respected: an ad can be punchier than a help article without being off-voice.
Every passage that departs from the baseline is marked with the rule it breaks and the evidence. Claims that exceed what the guide allows are flagged separately from tone issues.
Flagged passages are rewritten to pass, with meaning and length preserved. You accept, edit or reject each rewrite.
The voice profile is saved to your workspace, so future checks and future drafts by Sprites start from the same rules without re-uploading the guide.
Check a week of ads, emails and posts in one run and find out the paid social copy has drifted formal while the blog went casual.
Give new contributors a concrete, example-backed rule set and check their first drafts before they go anywhere near the site.
Treat the guide as a claims policy: flag promises, superlatives and comparisons the legal team has ruled out, across every channel.
Keep one voice profile per client workspace and check every deliverable against it before it reaches the client for approval.
From the material you give it: a style guide, approved examples, banned and required terms, or a plain description. It turns these into explicit rules it can cite in findings, so every flag points to the rule it breaks rather than a vague impression.
Yes. Paste or upload ads, emails, landing pages, posts and articles in one run. Sprites applies the same voice baseline while allowing for channel norms, so a short ad is not penalized for being short.
Vocabulary and banned terms, formality, point of view, sentence rhythm, reading level, how the brand and the customer are named, and factual or promotional claims that go past what the guide permits.
Both. Each flagged passage comes with a rewrite that keeps the meaning and fixes the voice. Accept it, tweak it or ask for a different take, and Sprites revises in the same thread.
Yes. Sprites has an official MCP server, so you can hand Claude, ChatGPT, Grok or any MCP-compatible assistant a draft and get the same voice check and rewrites, using the profile saved in your Sprites workspace.
Other prompts that run on the same agent. Browse the full gallery at /templates.
