SAPARI — Writing Guidelines¶

These rules apply to everything: social captions, blog posts, carousel copy, landing pages, emails, product UI, demo videos. If it has words and the Sapari name on it, these apply.

The voice¶

Write like a sharp friend who knows a lot about video and content creation. Direct, opinionated, experienced. Not a brand. Not a marketing team. Not an AI.

Use contractions naturally (you'll, don't, it's, we're). Be honest about trade-offs. Be opinionated, but show your reasoning. State things directly without preamble.

Who we're talking to¶

Creators who want their content to be good and don't want to spend their life in post-production to get there.

That includes the TikToker posting 5 clips a day who needs clean audio and captions on every one. The YouTuber who'd rather spend 20 minutes reviewing AI edits than 4 hours cutting silence manually. The course creator staring down 40 lectures that all need the same treatment. The coach who records great advice on their phone and wants it to sound like they have a production team.

They all share one thing: they care about how their content looks and sounds. They just refuse to let post-production eat their creative time.

We're not for people who don't care. If someone posts raw, unedited audio with mouth clicks and 30-second pauses and doesn't think twice about it — they won't see the value. That's fine. But everyone who's ever thought "I wish this sounded better" or "I don't have time to edit all of these" is exactly who Sapari is for.

No credential check. No subscriber count minimum. If you want your content to be good, you're in.

Banned phrases and patterns¶

These are the phrases that instantly make content sound like it was generated, not written. If you catch yourself typing any of these, stop and rewrite.

AI preamble and transitions¶

"Here's the thing:"
"Here's what's happening:"
"Here's why."
"Here's the deal:"
"Let's dive in"
"Let's break it down"
"Let's unpack this"
"Now let's look at..."
"Let's understand X"
"Without further ado"
"In this post, we'll explore..."
"So, what exactly is..."

Say the thing. No preamble, no announcements.

Artificial emphasis¶

"This is crucial/essential/vital/key"
"This is a game-changer"
"The beauty of this is..."
"What's really powerful here is..."
"Here's the key insight:"
"The magic happens when..."
"At the end of the day:"
"It's worth noting that..."
"Needless to say..."
"It goes without saying..."

If it's important, the reader will know from the content. You don't need to announce importance.

Marketing and hype¶

"Revolutionary" / "game-changing" / "cutting-edge"
"Powerful" / "robust" / "elegant" / "seamless"
"Unlock" / "unleash" / "supercharge" / "level up"
"Elevate your content"
"Take your content to the next level"
"Transform your workflow"
"In today's fast-paced digital landscape"
"The future of content creation"

These words have been drained of all meaning. They communicate nothing except "a brand is talking at you."

Hedging and filler¶

"It might be worth considering..."
"Perhaps" / "it's possible that"
"It's important to note that..."
"It should be mentioned that..."
"Arguably"
"Basically" / "essentially" / "fundamentally"
"Straightforward" / "simply" / "just" / "easy"
"Various" / "numerous" / "a wide range of"

Either commit to the statement or don't make it. Hedging makes you sound unsure. Filler makes you sound like you're padding word count.

Engagement bait¶

"Save this"
"Share this with someone who needs it"
"Send this to a creator who..."
"Follow for more"
"Drop a [emoji] if you agree"
"Tag someone who..."
"Link in bio"
"Let me know in the comments"
"What do you think? 👇"

This is the content equivalent of begging. If the content is worth saving, they'll save it. If it's worth sharing, they'll share it. Asking for it cheapens everything before it.

Rhetorical sales questions¶

"But what if there was a better way?"
"Sound familiar?"
"What if you could..."
"Want to know the secret?"
"Ready to transform your..."

Rhetorical questions that exist only to set up a pitch. The reader feels the turn coming and checks out.

Cliché transitions¶

"That being said..."
"With that in mind..."
"That said..."
"Moving on..."
"On that note..."
"Speaking of which..."
"But wait, there's more"

These are the verbal equivalent of clearing your throat. Cut them and go directly to the next point.

AI rhythm patterns¶

These are structural tells, not individual phrases. They're harder to spot because each sentence sounds fine alone. The pattern is the problem.

Short punchy sentence chains. When every sentence is 4-8 words and lands like a drum beat, it reads like a machine generating one confident statement after another. Real writing varies sentence length. A long thought followed by a short one has rhythm. Seven short ones in a row has a pattern.

Bad: "You have the ideas. You recorded the video. It's on your hard drive. But you haven't posted it. The editing takes too long. So it sits there. And nothing happens."

Good: "You recorded the video three days ago and it's still sitting on your hard drive, not because it's bad but because there's two hours of captioning and exporting between you and actually posting it."

The bad version has 7 sentences averaging 6 words each. The good version has one sentence that breathes like someone actually talking.

The correction structure: "Not because X. Because Y."

"Not because it's hard. Because it's tedious."
"Not because they don't care. Because they don't have time."
"It's not about quality. It's about consistency."

This is one of the strongest AI tells in 2025/2026. The pattern is always: negate the obvious assumption, then state the real reason in a parallel structure. Real people occasionally do this. AI does it in every other paragraph. If you catch yourself writing "Not because... Because..." or "It's not X. It's Y." more than once per post, rewrite at least one of them into a normal sentence.

The staccato qualification: "Not X. Not Y. But Z."

"Not for everything. Not for everyone. But for most creators? Yes."
"Not a studio. Not a $400 interface. A USB mic."
"Not slowly. Not gradually. Immediately."

Three beats, escalating. The third breaks the pattern. This is a rhetorical device that AI uses constantly because it creates a sense of building momentum with minimal effort. Once per post maximum, and only if the rhythm genuinely serves the point. If you find yourself reaching for it, write a normal sentence instead.

"Most creators" and other vague majorities.

"Most creators think..."
"Many people don't realize..."
"A lot of creators struggle with..."
"So many people..."

These generalize without evidence and set up a "but actually" correction. If you mean "you," say "you." If you mean a specific type of creator, name them. "Most creators" is the social media equivalent of "studies show."

Bad: "Most creators think their problem is perfectionism." Good: "You keep telling yourself it's perfectionism."

What to do instead¶

State things directly¶

Bad: "Here's the thing: your audience can tell when you're being authentic." Good: "Your audience can tell when you're being authentic."

Bad: "It's worth noting that 62% of readers prefer human content." Good: "62% of readers prefer human content."

Bad: "Let's dive into why pacing matters." Good: "Pacing matters because..."

The fix is almost always: delete the preamble, keep the statement.

Lead with the point¶

Bad: "In today's creator economy, many creators struggle with the challenge of scaling their content output while maintaining the quality that their audience has come to expect." Good: "Scaling breaks your quality. Or it used to."

Bad: "If you've ever found yourself spending hours on post-production, you're not alone." Good: "You spend more time in post than in production. The ratio is broken."

Show your reasoning¶

Bad: "Pacing is essential for great content." Good: "Pacing controls whether your audience finishes the video or drops off at the 30-second mark. It's the difference between a viewer who watches and a viewer who watched."

Bad: "You should post consistently." Good: "208 experiments confirmed that repeated exposure increases preference. Your audience literally trusts you more just because they keep seeing you. Consistency isn't a growth hack. It's neuroscience."

Be specific, not general¶

Bad: "Great storytellers understand their audience." Good: "When a story works, the listener's brain activity mirrors the speaker's with a 1-3 second delay. In some regions, it gets ahead. Princeton measured this."

Bad: "Quality content performs better." Good: "Sam Sulek gained millions of subscribers posting videos with virtually no editing. The market is self-correcting."

Use honest transitions¶

Bad: "Now let's look at why this matters for your content strategy." Good: "But that only explains why people click. It doesn't explain why they stay."

Bad: "Let's dive into the science behind engagement." Good: "The science is specific about this."

Bad: "Moving on to our next point..." Good: [Start the next point. No announcement needed.]

The best transition is a question the previous section raised, or a gap the reader can feel.

Writing about the product¶

When writing about what Sapari does, stay concrete. Describe the action and the result, not the category.

Bad: "Sapari's AI-powered editing suite streamlines your post-production workflow." Good: "Upload a video. Sapari cuts the silence, removes false starts, and adds captions. You review, tweak, export."

Bad: "Our intelligent silence detection leverages advanced transcription technology." Good: "Sapari listens to your audio, finds the pauses, and cuts them. You control how aggressive the cuts are with a slider."

Bad: "Flexible credit-based pricing that scales with your needs." Good: "You only pay for the AI features you use. Silence removal and captions cost 0.3 credits per minute. The full pipeline costs 1 credit. Manual editing and rendering are free."

When comparing to competitors, don't name them or trash them. Describe the problem with the alternative approach.

Bad: "Unlike Descript, Sapari doesn't force you into a text-based editing paradigm." Good: "Most AI editors make you learn a new way to edit. Sapari gives you a timeline. If you've used any video editor, you already know how it works."

When describing speed or volume, use real numbers.

Bad: "Edit videos faster than ever before." Good: "A 10-minute video takes about 3 minutes to analyze. A 2-minute TikTok takes 40 seconds."

Writing for different segments¶

The voice stays the same. The examples and pain points shift.

Short-form creators (TikTok, Reels, Shorts): Lead with volume and speed. They know the editing-per-clip problem intimately. "You shot 8 clips today. Editing all 8 in CapCut will take you an hour. Sapari does it in the time it takes you to get coffee."

Long-form creators (YouTube, podcasts): Lead with the tedium of long edits and the quality of AI detection. "Your 45-minute recording has 6 minutes of dead air and 23 false starts. You could find them manually, or you could let Sapari flag them in 4 minutes and spend your time on the creative decisions."

Course and education creators: Lead with batch processing and consistency. "40 lectures. Same intro, same outro, same caption style, same silence threshold. Set it once, apply it to all of them."

Business and professional (LinkedIn, coaches, consultants): Lead with looking professional without the overhead. "You recorded good advice on your phone. Sapari makes it sound like you have a production team. Your audience won't know the difference."

Punctuation¶

Colons for definitions and explanations: "The formula is simple: tension sustained through narrative produces behavior change."

Parentheses for asides and examples: "This works across every format (YouTube, newsletters, TikTok, podcasts)."

Periods to separate thoughts. Not commas. Not em dashes.

Bad: "The hardware doesn't matter, what makes it work is the story." Good: "The hardware doesn't matter. What makes it work is the story."

Em dashes: Use sparingly. One per post maximum. When in doubt, use a colon, parentheses, or split into two sentences. Em dash overuse is one of the strongest AI tells in 2025/2026 writing.

Lowercase or sentence case. The UPPERCASE is for the graphics, not the caption.
Short paragraphs. One idea per paragraph.
No emoji (or one, maximum, if it genuinely adds something).
No hashtag walls. Zero to three hashtags, placed naturally or at the end.
No bullet lists in captions. Write in sentences.
Line breaks between paragraphs for readability.

Formatting in blog posts¶

Short paragraphs. 2-4 sentences.
No headers that sound like AI wrote them ("Why This Matters", "The Bottom Line", "Key Takeaways", "Putting It All Together").
Section headers should be direct and descriptive: "Neural coupling", "The overstim era died in 2024", "Specificity is the soul of narrative."
Bold for terms on first use, not for emphasis throughout.
No emoji.
Bullet lists only for genuine reference material. Use prose to explain concepts.

Formatting in demo content (TikTok, Reels, Shorts)¶

Demo content is where the product sells itself. The video IS the proof.

Screen recordings over talking head. Show the timeline, the AI detections, the before/after.
Caption the demo video using Sapari. Meta-demonstration.
Keep it under 60 seconds. One feature per video. "Watch Sapari remove 47 silences from this clip in 30 seconds."
Voiceover in the Sapari voice: direct, no hype, describe what's happening. "This is 12 minutes of raw audio. 3 minutes of it is dead air. Gone."
End on the result, not a CTA. Show the clean waveform, the tight edit, the final export. Let the quality speak.

Citing research¶

Sapari content cites real research. That's a core part of the brand. But citing it wrong is worse than not citing it at all.

Every claim needs a verifiable source¶

If you write "208 experiments confirmed that familiarity breeds preference," you need to link to the actual meta-analysis. Not to a blog post that mentions it. Not to a Wikipedia summary. The original paper, or at minimum a direct institutional source.

If you can't find the original source, don't cite the claim. Paraphrase it as general knowledge or drop it entirely.

Always verify before publishing¶

Research gets misquoted constantly. The telephone game from paper to pop-science article to blog post to social media post degrades accuracy at every step. Before using any stat or finding:

Find the original paper or primary source.
Read the relevant section yourself (not just the abstract).
Confirm the claim actually says what you think it says.
Check the context. A finding about college students in a lab doesn't automatically generalize to all humans everywhere.

How to cite in different formats¶

Blog posts: Hyperlink the researcher's name or the study title to the source. First mention gets the full context: "Uri Hasson's neural coupling research at Princeton" linked to the paper or the Princeton page. Don't write "studies show."

Social captions: Name the researcher and institution. You can't hyperlink in Instagram captions, so be specific enough that someone could find it: "Paul Zak's lab at Claremont Graduate University" not "a study found." If there's a well-known talk or article, mention it by name.

Carousels: Researcher name and institution on the slide itself. Keep it short but findable. "URI HASSON — PRINCETON" or "PAUL ZAK — CLAREMONT UNIVERSITY" in monospace under the claim.

Demo videos: No research citations needed. The product is the evidence. Show, don't cite.

What "studies show" actually means¶

Never write "studies show" or "research suggests" without attribution. It's the fastest way to sound like every other content account that half-read a blog post. Name the researcher, name the institution, name the year if relevant. Specificity is the difference between authority and hand-waving.

Bad: "Studies show that stories trigger oxytocin release." Good: "Paul Zak's neuroeconomics lab at Claremont found that character-driven stories with tension arcs trigger oxytocin synthesis. Participants given oxytocin donated to 57% more causes."

Bad: "Research suggests that surprise helps learning." Good: "A 2024 study in Nature Communications (Pine et al.) found that learning is strongest when new information contradicts existing beliefs, generating large prediction errors."

When you're wrong, correct it¶

If a published post contains a misquoted stat or a broken citation, fix it immediately. Don't quietly edit. Add a note: "Corrected: original post cited X, actual finding is Y." Transparency builds more trust than pretending it didn't happen.

The test¶

Before publishing anything, read it aloud. If any sentence sounds like it came from a LinkedIn influencer, a marketing brief, or a ChatGPT output, rewrite it.

Three questions:

Would a real person actually say this to a friend?
Does every sentence contain actual information, or is it connecting tissue?
If you removed the Sapari name, could this have been written by any brand? If yes, it's not specific enough.

If the writing is good, you don't need to tell people it's good. If the point is sharp, you don't need to announce it's sharp. Say the thing. Trust the reader.