电影感产品发布影片

You are an elite creative director, motion designer, product filmmaker, frontend engineer, and cinematic storyteller.

Your task is to CREATE A COMPLETE, POLISHED, PRODUCTION-READY GTM LAUNCH VIDEO FOR:

# CLAUDE SONNET 5.5

This should NOT feel like a generic AI explainer, SaaS promo, Canva presentation, or slideshow.

It should feel like a **premium cinematic technology launch film** produced by a world-class studio.

Think:

* Apple product launch film
* high-end AI research visualization
* cinematic title sequence
* sophisticated motion graphics
* premium brand campaign
* futuristic but restrained
* extremely polished
* visually mesmerizing
* editorial typography
* dramatic pacing
* exceptional sound-design thinking

The viewer should finish the video thinking:

**“This is not just a faster model. This is a fundamentally more capable everyday AI system.”**

---

# 1. CORE MESSAGE

The central narrative:

**Claude Sonnet 5.5 takes the intelligence of the Claude 5.5 generation and makes it faster, more efficient, and more practical for everyday work.**

The story should revolve around three ideas:

**INTELLIGENCE**
**SPEED**
**EFFICIENCY**

Then demonstrate what those mean in the real world:

**Code.**
**Reasoning.**
**Computer use.**
**Visual understanding.**
**Knowledge work.**
**Design.**
**Collaboration.**

Do NOT dump features onto the screen.

Instead, make the viewer EXPERIENCE the capabilities.

---

# 2. LENGTH

Create a cinematic film approximately:

**75–100 seconds**

The pacing should feel intentional.

First 10 seconds:
HOOK.

10–25 seconds:
INTRODUCE SONNET 5.5.

25–65 seconds:
CAPABILITY JOURNEY.

65–80 seconds:
SPEED + COST.

80–95 seconds:
POSITIONING + ENDING.

Do not make every section the same length.

Use pauses and negative space.

Let important statements breathe.

---

# 3. OPENING — THE HOOK

Do NOT begin with:

“Introducing Claude Sonnet 5.5”

That is boring.

Begin with a visual metaphor.

Black screen.

Almost total silence.

A tiny point of light appears.

It begins processing.

The point becomes thousands of rapidly moving particles.

The particles form fragments of:

code,
documents,
charts,
UI,
terminal commands,
images,
graphs,
text.

Everything moves extremely fast.

Then—

**silence.**

Everything freezes.

A single sentence appears:

> “What happens when intelligence gets faster?”

Pause.

Then:

# CLAUDE SONNET 5.5

Minimal.

Beautiful.

Large typography.

Subtle motion.

No cheesy effects.

---

# 4. PRODUCT POSITIONING

Introduce Sonnet 5.5 as:

**A faster, lower-cost complement to Claude Opus 5.5.**

Visually establish the Claude 5.5 family:

OPUS 5.5
SONNET 5.5
HAIKU 5.5

But Sonnet should become the visual focus.

Communicate that:

Opus 5.5 → complex work requiring sustained judgment

Sonnet 5.5 → well-scoped everyday tasks, fixing bugs, creating polished documents, slides and spreadsheets

Haiku 5.5 → high-volume, cost-sensitive applications

Do not spend too much time explaining this.

Use elegant visual hierarchy instead.

Source: Anthropic describes Sonnet 5.5 as the second model in the Claude 5.5 family and positions it as a faster, lower-cost complement to Opus 5.5.

---

# 5. THE BIG PERFORMANCE REVEAL

Create a dramatic data visualization sequence.

Start with:

SONNET 5

Then introduce:

SONNET 5.5

Show Terminal-Bench 4.0:

**10.3% → 70.6%**

Do not simply display a bar chart.

Make the number physically transform.

10.3%

The number fractures.

Explodes into particles.

Reassembles as:

70.6%

Then cut directly into visual representations of an agent working inside a terminal.

Terminal commands appearing.

Files changing.

Tests running.

Code being inspected.

Errors appearing.

Claude fixing them.

The system keeps moving.

Voiceover / typography:

**“From 10.3% to 70.6% on Terminal-Bench 4.0.”**

Then:

**“Agentic coding performance, dramatically upgraded.”**

Important:
Do not claim Sonnet 5.5 is universally superior to every model.

The source explicitly notes that benchmark scores capture only one facet of capability and that Opus 5.5 remains stronger for complex, open-ended work requiring sustained judgment.

---

# 6. CODING SEQUENCE

This should be one of the hero sequences.

Visual language:

A developer's environment.

Dark terminal.

Code.

Git diff.

Architecture diagrams.

Test results.

Logs.

Multiple windows.

Instead of showing static screenshots, create the feeling of an AI actually working.

Show:

READ

UNDERSTAND

PLAN

EDIT

RUN

TEST

FIX

SHIP

These words appear briefly as kinetic typography.

Then accelerate everything.

The interface begins moving faster.

The camera pushes through the screen.

We transition from terminal → code → architecture → finished application.

On-screen statement:

**“Understand the codebase.
Make the change.
Check the work.”**

Mention that early testers reported Sonnet 5.5 understood codebases quickly and often batched tool calls into fewer steps, reducing cost.

---

# 7. KNOWLEDGE WORK

Transition away from code.

A large pile of information appears:

PDFs

spreadsheets

earnings reports

transcripts

charts

documents

slides

The information overwhelms the screen.

Then Sonnet cuts through the noise.

Everything gets organized.

Information flows into:

ANALYZE

SYNTHESIZE

STRUCTURE

CREATE

The output becomes a beautifully designed presentation.

Show a polished operating-review deck being assembled.

Use the Anthropic example:

A public company's quarterly earnings materials + call transcripts + slide template

→

10-slide operating review.

The source says two experts judged one internal test's first draft ready to send as-is. Present this as an example, not a universal guarantee.

---

# 8. VISUAL UNDERSTANDING

Create an elegant sequence involving charts and images.

Start with a complicated chart.

Camera moves toward it.

The chart breaks apart into visual components.

Sonnet identifies relationships.

Numbers transform into explanations.

The visualization becomes understandable.

Use:

**“See the information.”**

Then:

**“Understand the information.”**

Then:

**“Act on it.”**

Reference the published Chartography score:

**61.6%**

versus

**15.6%**

for Sonnet 5.

Also show:

**80.1% OSWorld**

and the broader computer-use story.

Keep benchmark visuals tasteful.

No giant dashboard.

No spreadsheet aesthetic.

---

# 9. LONG-HORIZON / UNEXPECTED CAPABILITY MOMENT

This should be one of the film's most memorable moments.

Introduce:

**Pokémon Red**

But do it cinematically.

Old-school pixel aesthetic appears.

The camera flies into the game world.

A screenshot becomes the environment.

Sonnet interprets what it sees.

Moves.

Adapts.

Continues.

Then pull back from the pixel world into modern AI visuals.

Typography:

**“The first Sonnet model to beat Pokémon Red using only screenshots.”**

This is directly supported by Anthropic's launch page.

---

# 10. SPEED

Now change the energy of the entire film.

Everything becomes faster.

Cursor.

Terminal.

Documents.

Animations.

Responses.

Tool calls.

Words.

Then everything stops.

Huge typography:

# 30%+

A second line:

**FASTER OUTPUTS**

Then:

**Sonnet 5.5 generates outputs more than 30% faster than Sonnet 5.**

This is Anthropic's published positioning.

Use visual motion to communicate speed instead of relying on narration.

---

# 11. COST / EFFICIENCY

This should feel like the “surprise” section.

Show two parallel paths.

SONNET 5

Longer path
More tokens
More steps
Higher task cost

SONNET 5.5

Shorter path
Fewer tokens
Fewer steps
Lower task cost

Then merge the two paths into a single number:

**UP TO 30% LESS PER TASK**

The pricing should appear separately:

INPUT

**$2 / 1M tokens**

OUTPUT

**$10 / 1M tokens**

CACHE READS

**$0.20 / 1M tokens**

CACHE WRITES

**$2.50 / 1M tokens**

Do not imply the list price itself was reduced versus Sonnet 5.

The key point is task efficiency.

## Anthropic states the token prices are the same as Sonnet 5, while Sonnet 5.5 typically requires fewer tokens per task and costs up to 30% less per task in testing.

# 12. EFFORT LEVEL

Create a beautiful visual metaphor for effort.

A single Claude interface.

A slider moves:

LOW → MEDIUM → HIGH

At LOW:

FAST

At HIGH:

DEEPER REASONING

Communicate:

**Choose how much effort Claude spends.**

Show the tradeoff:

Speed ↔ Cost ↔ Quality

Anthropic notes that lower effort produces faster, lower-token responses, while higher effort allows longer reasoning and more checking.

Keep this section under 6 seconds.

---

# 13. COLLABORATION

Show a person working with Claude.

Not “human versus AI.”

Instead:

human → AI → human → AI

Rapid iteration.

A rough idea becomes:

code

document

spreadsheet

presentation

design

finished product

Use the message:

**“Better collaboration.
Faster iteration.”**

Anthropic says early testers described Sonnet 5.5 as a better collaboration partner than Sonnet 5, with speed supporting rapid iteration on less complex tasks.

---

# 14. SAFETY

Do NOT turn this into a scary cybersecurity montage.

Keep it elegant.

Show the system approaching a boundary.

A glowing perimeter appears.

Certain high-risk paths are blocked.

Routine work continues uninterrupted.

Typography:

**CAPABLE**

then:

**CONTROLLED**

Then:

**BUILT WITH SAFEGUARDS FOR HIGH-RISK USE**

Communicate that Sonnet 5.5 launches with cyber safeguards and fallbacks similar to those developed for more capable models.

Do not overstate this.

## Routine software development and most life-sciences work are described by Anthropic as unaffected by these safeguards.

# 15. FINAL SEQUENCE

Return to the original point of light.

But now it is enormous.

It contains everything we saw:

code

documents

charts

images

interfaces

terminal

research

design

All orbiting around one point.

The system contracts into a single elegant symbol.

Silence.

Then:

# CLAUDE SONNET 5.5

Underneath:

**Faster.
More capable.
More efficient.**

Then one final line:

**Built for the work between the big breakthroughs.**

Hold.

Fade to black.

Anthropic logo / Claude wordmark treatment.

End.

---

# 16. VISUAL DIRECTION

The visual quality is critical.

Use:

* cinematic depth
* volumetric light
* subtle particles
* glass interfaces
* realistic screen reflections
* macro details
* sophisticated typography
* smooth camera movement
* 3D space
* depth of field
* subtle grain
* elegant transitions
* premium UI animation
* kinetic typography
* data visualizations
* procedural motion
* intentional negative space

Avoid:

* cheesy AI imagery
* robots
* glowing brains
* generic humanoid AI
* stock footage
* cliché “future city” visuals
* excessive neon
* overused holograms
* template-style transitions
* random particles everywhere
* corporate PowerPoint aesthetics
* excessive text

The visual language should feel **intelligent, restrained, premium and modern.**

---

# 17. COLOR / BRAND

Use a restrained Anthropic-inspired palette.

Primarily:

deep blacks
warm whites
soft neutrals
subtle Claude orange accents

Orange should be used strategically.

Do not turn the entire video orange.

Typography should be editorial and premium.

Use extremely large type for important statements.

Use small precise typography for benchmark data.

---

# 18. SOUND DESIGN

Treat sound as a major part of the experience.

Create a conceptual soundscape that evolves with the intelligence of the system.

Beginning:

near silence

low-frequency ambience

tiny digital clicks

Middle:

keyboard sounds

terminal activity

subtle system pulses

data movement

rising tension

Speed section:

rapid rhythmic elements

compressed transients

increasing tempo

Final reveal:

everything drops away

single tonal hit

silence

then the final title.

The audio should feel like a film, not background music.

---

# 19. EDITING LANGUAGE

Use:

hard cuts where impact is needed

slow cinematic transitions where context matters

match cuts

particle transformations

camera push-ins

screen-to-world transitions

typographic wipes

data morphing

light-based transitions

motion blur

depth transitions

Do NOT over-transition.

Every transition should communicate a conceptual connection.

---

# 20. BENCHMARKS TO FEATURE

Prioritize the benchmarks visually instead of showing every benchmark.

Use:

Terminal-Bench 4.0
70.6%

Sonnet 5:
10.3%

FrontierCode:
46.2% Max

CursorBench:
55.5%

GDPval-AA:
1844

AA-Briefcase:
1811

Humanity's Last Exam:
64.5% with tools

OSWorld:
80.1% partial

Chartography:
61.6% no tools

These figures come from Anthropic's published benchmark table.

Do not misleadingly imply that one benchmark proves overall intelligence.

---

# 21. ACCURACY RULE

This is extremely important.

Do NOT invent:

benchmarks

customers

quotes

performance numbers

capabilities

pricing

features

comparisons

scientific claims

Do not make unsupported claims like:

“best model in the world”

“beats every model”

“AGI”

“human-level”

“revolutionary”

“the smartest AI ever”

unless explicitly supported by the provided source.

Use the supplied Anthropic source as the factual source of truth.

---

# 22. PRODUCTION QUALITY

Do NOT simply produce a storyboard.

Actually build the complete experience.

If this is being implemented as code:

1. Inspect the existing project.
2. Determine the best technology for cinematic motion.
3. Build reusable animation primitives.
4. Build the timeline as modular scenes.
5. Keep typography perfectly synchronized.
6. Make every animation deterministic.
7. Optimize rendering performance.
8. Avoid unnecessarily heavy assets.
9. Ensure the result runs smoothly.
10. Make the entire film feel authored rather than procedurally random.

The final result should feel like an actual professional launch film.

---

# 23. RESPONSIVENESS / EXPORT

Design the composition primarily for:

**1920 × 1080**

16:9 cinematic master.

But structure the scene system so it can later be adapted to:

9:16

1:1

without rewriting the entire animation.

---

# 24. IMPORTANT CREATIVE PRINCIPLE

Do not explain AI.

**SHOW AI WORKING.**

Instead of saying:

“Sonnet 5.5 is good at coding.”

Show:

problem → codebase → reasoning → changes → tests → working result.

Instead of:

“Sonnet 5.5 is good at knowledge work.”

Show:

chaotic information → understanding → synthesis → polished deliverable.

Instead of:

“Sonnet is faster.”

Show the entire world accelerating.

Make the audience FEEL the improvement.

---

# 25. NARRATION

Use extremely sparse narration.

The film should work even with the sound off.

Narration should feel like a premium technology documentary.

Short sentences.

Confident.

No marketing fluff.

Example style:

“What happens when intelligence gets faster?”

“Code.”

“Reasoning.”

“Understanding.”

“Creation.”

“Then everything moves faster.”

“Same model family.”

“Less waiting.”

“Fewer tokens.”

“More work done.”

“Claude Sonnet 5.5.”

Do NOT narrate every fact.

Let visuals carry the story.

---

# 26. FINAL EXPERIENCE

The finished result should have:

**A killer first 5 seconds.**

**A memorable middle.**

**A visually spectacular climax.**

**A clean final brand lockup.**

It should be good enough to publish as:

* an Anthropic-style launch film
* X/Twitter launch content
* a website hero video
* Product Hunt launch material
* developer marketing
* conference screens
* YouTube

The viewer should understand three things by the end:

### 1.

**Sonnet 5.5 is substantially more capable than Sonnet 5 in many evaluated tasks.**

### 2.

**It is significantly faster and more efficient.**

### 3.

**It is designed for practical, everyday professional work.**

Build the experience around those three ideas.

Do not make this feel like an advertisement.

Make it feel like **the arrival of a new piece of technology.**

出处：作者在原帖下的回复。提示词保留原文语言，未做改写。

更多产品广告案例

• 1:07

  云端AI平台3D游戏生成演示

  431.6万 播放

• 0:24

  可实拼真人尺寸乐高鸭设计

  132.3万 播放

• 0:10

  After Effects制作的产品发布视频

  75.4万 播放

• 1:02

  东京末班车地图互动网站

  72.3万 播放

• 0:41

  管理编程 Agent 的虚拟办公室

  47.1万 播放

• 0:31

  可交互动态演示幻灯片

  44万 播放