⚡ Promptolis Original · Life & Relationships
🥂 Wedding Speech Architect
The opening that isn't 'I've known her since...', the anecdote that lands, and the line that makes Mom cry.
Why this is epic
Refuses the four openings every wedding guest has heard 400 times ('I've known X since...', 'When Sarah first told me about...', 'For those who don't know me...', 'They say marriage is...') and gives you one nobody has used.
Engineers two anecdotes using the 'specific-detail-then-meaning' structure — not 'he's a great guy' vague — so the room actually sees the person.
Identifies the single emotional sentence (we call it The Cry Line) placed at the 75% mark, because that's where the room's emotional peak naturally lands.
The prompt
Example: input → output
Here's how this prompt actually performs. Real input below, real output from Claude Opus 4.
Speaker's role: Maid of Honor Speaker's name: Priya Couple's names: Hannah and Marcus Which one you know primarily: Hannah — we were roommates at Northwestern freshman year (2014) and lived together through 2019. Relationship length and nature: 11 years. She's the person I call first for everything. I was there when she met Marcus in 2021 on a hiking trip in Colorado. Tone preference: Balanced — maybe 60% warm, 40% funny. I don't want to roast her but I want people to laugh twice. 3-5 specific memories: - Freshman year, she tried to make Diwali sweets for our dorm floor, set off the fire alarm at 11pm, and when the RA came she offered him a burned gulab jamun and he actually ate it. - 2018, her dad was diagnosed with cancer (he's fine now, in remission) and she drove home to Minneapolis every other weekend for a year. Never missed a work deadline. I didn't know until month 4. - The night she met Marcus, she called me from a gas station in Boulder at 1am and said 'I think I just met someone who listens on purpose.' That phrase stuck with me. - Last year, Marcus learned to make her mom's chicken curry recipe and FaceTimed her mom to check the spice level. Hannah cried when she told me about it. What you love about the partner: Marcus is quiet but deeply attentive. He remembers things. He makes Hannah slower, in a good way — she used to be a tornado, now she's a river. Your biggest fear: I will cry before I finish. I always cry. Any landmines: Hannah's dad is here and it's emotional for the family given his cancer history. Marcus's parents are divorced but both remarried amicably — all four are here. Wedding vibe: Semi-formal, outdoor ceremony in Napa.
Wedding Speech Architect — Priya for Hannah & Marcus
A wedding speech is a 3-4 minute oral essay that a room of 80-200 people will remember for exactly one of two reasons: it was specific, or it was cringe. In our experience editing 200+ speeches, the difference is almost never talent — it's structure. Below is yours, engineered around the 'specific-detail-then-meaning' method popularized in narrative nonfiction (and borrowed from classic Aristotelian ethos-pathos-logos pacing).
Speech Architecture (why this works)
| Element | Choice | Why |
|---|---|---|
| Opening | Cold-open on the gas station call | Avoids the 4 dead openings; drops the room into a specific moment within 8 seconds |
| Anecdote 1 | Diwali fire alarm (2014) | Proves Hannah's warmth + humor. Gets the first laugh at ~45 seconds, which is the ideal position |
| Anecdote 2 | Marcus learning the curry recipe | Proves who he is without him saying a word. Bridges to the Cry Line |
| The Cry Line | 'He didn't learn the recipe to impress her. He learned it because he saw how she looked when she missed home.' | Placed at the 75% mark. Voice drops. 2-second pause after. |
| Close | Toast framed around 'the person who listens on purpose' — callback to opening | Circular structure. Clean lift. No wobble. |
Total length: 512 words, ~3 minutes 20 seconds spoken at a normal pace (roughly 155 words per minute, which is the sweet spot — faster and it feels rushed, slower and you lose the room).
The Speech (512 words / ~3 min 20 sec)
At 1 a.m. on a Tuesday in October 2021, my phone rang from a gas station in Boulder, Colorado. [pause] Hannah had just met Marcus that afternoon on a hiking trail. And the first thing she said to me — not 'he's cute,' not 'he's funny' — she said: [slow] 'I think I just met someone who listens on purpose.'
I wrote that sentence down. I've been waiting four years to say it out loud.
[beat]
Hi, I'm Priya. Hannah and I have been roommates, co-conspirators, and each other's 2 a.m. phone call since freshman year of Northwestern. Which is also, not coincidentally, the year she tried to make Diwali sweets for our entire dorm floor, set off the fire alarm at 11 p.m., and — when the RA came upstairs furious — offered him a slightly charred gulab jamun. [pause for laugh] He ate it. He ATE the burned one. That is the Hannah effect. People eat the burned gulab jamun, and then they thank her.
[beat]
But here is what most of you don't know about Hannah. In 2018, her dad got sick. [soft] And for a year, she drove home to Minneapolis every other weekend. Six hours each way. She didn't miss a single work deadline. She didn't tell me for four months. [pause] That's who she is. She carries the heavy things quietly, and she shows up with sweets.
Which brings me to Marcus. [look at Marcus]
I have watched Hannah fall for exactly one person in eleven years, and I have paid close attention. Here is what I noticed. Last year, Marcus learned how to make Hannah's mom's chicken curry. From scratch. [pause] And he FaceTimed her mother — in Minneapolis — to check the spice level.
[voice drops, slower]
He didn't learn the recipe to impress her. He learned it because he saw how she looked when she missed home. [2-second pause]
That is the man my best friend is marrying.
[beat, lift voice]
Hannah — you used to be a tornado. Beautiful, but, you know, a tornado. [small laugh] And somewhere in the last four years, you became a river. Still going somewhere. Just — steadier. Marcus did that. You did that together.
[raise glass, look at the room]
So. To Hannah and Marcus. To the burned gulab jamun and the chicken curry. To the dads who fought to be here tonight [nod toward Hannah's father], and to the people who listen on purpose.
May your marriage be the kind where someone always learns the recipe.
[to the couple] I love you both. Cheers.
Delivery Notes
- Pause for laughs: After 'He ATE the burned one' (1.5 seconds minimum — don't step on the laugh). After 'people eat the burned gulab jamun, and then they thank her' (shorter, 1 second).
- Voice drops at: 'He didn't learn the recipe to impress her.' This is the Cry Line. Slow down 30%. Do NOT try to hold eye contact with Hannah here — you said you'll cry. Look at a fixed point on the back wall.
- Look at the couple during: the Marcus segment, and the final 'I love you both.' Look at the room during the toast lift.
- The one line to cut if you're running long: 'Beautiful, but, you know, a tornado.' It's the weakest laugh and the speech survives without it.
What About the Landmines?
- Hannah's dad: The line 'the dads who fought to be here tonight' acknowledges him with dignity without making the speech about his illness. It will mean something to the family and wash over everyone else. If he's too emotional about it, change to 'the parents who got us here.'
- Marcus's divorced parents (all 4 present): The speech never references his family structure, which is correct. Don't try to name-check all four — you'll get it wrong and it'll feel transactional.
- Your crying: The Cry Line is engineered so that if you cry there, it's on-brand, not off-script. The room will cry with you. The danger zone is the opening — if you cry in the first 30 seconds, you lose the room. Breathe before 'At 1 a.m. on a Tuesday.'
Key Takeaways
- Open cold on a specific moment, not an introduction. 'At 1 a.m. on a Tuesday' beats 'I've known Hannah since 2014' every time.
- Two anecdotes, not three. One proves the person, one proves the couple. Three is a speech; two is a story.
- The Cry Line goes at 75%, not the end. The end is a toast. Toasts should lift, not sob.
- Specific details are the whole game. 'Burned gulab jamun,' 'six hours each way,' 'FaceTimed her mother' — these are what the room remembers in 10 years, not adjectives.
- Read it aloud once. Anything you stumble on, cut. If you can't say it smoothly in your living room, you won't say it smoothly at the altar.
If You Hate a Section
Tell me which paragraph isn't you. The tornado-to-river line is the most 'writerly' moment in the speech — if it feels too literary coming out of your mouth, I'll swap it for something plainer. Same function, different words.
Common use cases
- Best man speech with humor that doesn't roast the groom into a divorce
- Maid of honor speech balancing funny college stories with 'I'm proud of you'
- Father-of-the-bride speech that avoids the three dad-speech clichés
- Mother-of-the-groom speech welcoming the new daughter-in-law without being weird about it
- Sibling speech when you genuinely don't love your new in-law (tactful version)
- Rehearsal dinner toast (shorter, 90-second variant)
- Second-marriage speech where kids from prior marriages are present
Best AI model for this
Claude Sonnet 4.5 or GPT-5. Claude handles the emotional register better — it resists the 'cheesy wedding cliché' default that GPT occasionally slides into. For pure humor-forward best man speeches, GPT-5 edges ahead.
Pro tips
- Paste in 3-5 actual specific memories with details (the bar, the year, the exact thing they said). Vague inputs = vague speeches. The architect can only be as specific as you are.
- Tell it your ONE fear about giving this speech ('I'll cry', 'I'll be too sarcastic', 'I barely know the bride'). It rewrites around your failure mode.
- Ask for the 'read-aloud timing version' — it adds [pause], [look at couple], [wait for laugh] stage directions.
- If you hate a line, paste it back and say 'replace this — same function, different words.' Don't regenerate the whole speech.
- Read it out loud once before approving. Anything you stumble on, cut. Spoken English ≠ written English.
- The Cry Line only works if it's TRUE. If the architect writes something that isn't quite right, tell it the real thing and let it rephrase.
Customization tips
- Feed it REAL specifics. The burned gulab jamun detail is what makes this speech. 'She's always been generous' would have produced a worse speech. Give the architect the specific bar, the specific year, the exact phrase someone said.
- Tell it your failure mode. Priya said 'I always cry' — notice the architect placed the Cry Line strategically AND told her where to look during it. Your fear shapes the structure.
- Ask for the 'timing version' if you want stage directions in brackets ([pause], [look at couple]) embedded throughout. Some speakers love these, some find them distracting.
- For roast-heavy best man energy, tell it: 'I want three jokes that punch sideways, not down. Test each one against: would the groom's grandma laugh?' Claude is good at this filter.
- If the speech doesn't sound like you, paste back one paragraph and say 'rewrite this in my voice — I talk more like [X].' Don't regenerate the whole thing; you'll lose the good parts.
Variants
The 90-Second Toast
Shorter rehearsal-dinner or ceremony version. One anecdote, no roast, high-emotion close.
The Roast-Heavy Best Man
Leans 70% humor, 30% sentiment. Includes three roast-lines with 'safety checks' (is this going to land with grandma).
The Blended-Family Edition
For second marriages, step-parents, or speeches where ex-spouses/kids-from-prior-marriages are in the room. Navigates the complexity.
Frequently asked questions
How do I use the Wedding Speech Architect prompt?
Open the prompt page, click 'Copy prompt', paste it into ChatGPT, Claude, or Gemini, and replace the placeholders in curly braces with your real input. The prompt is also launchable directly in each model with one click.
Which AI model works best with Wedding Speech Architect?
Claude Sonnet 4.5 or GPT-5. Claude handles the emotional register better — it resists the 'cheesy wedding cliché' default that GPT occasionally slides into. For pure humor-forward best man speeches, GPT-5 edges ahead.
Can I customize the Wedding Speech Architect prompt for my use case?
Yes — every Promptolis Original is designed to be customized. Key levers: Paste in 3-5 actual specific memories with details (the bar, the year, the exact thing they said). Vague inputs = vague speeches. The architect can only be as specific as you are.; Tell it your ONE fear about giving this speech ('I'll cry', 'I'll be too sarcastic', 'I barely know the bride'). It rewrites around your failure mode.
Explore more Originals
Hand-crafted 2026-grade prompts that actually change how you work.
← All Promptolis Originals