Launch Videos

Today, we are releasing our state-of-the-art speech-to-text model called Willow Atlas One.

The spoken opening line (at most 25 words).

838k views

More views than 66% of the library · #203 of 598

Runtime
1:23
Posted
Apr 1, 2026
Pace
155 wpm
Willow on X

Introducing Atlas 1. Willow's new frontier speech-to-text model. It outperforms…

1:23 · Apr 1, 2026

Nothing from X loads until you click. Loading fetches X’s embed script, which can set cookies. Nothing autoplays from this page. Open on X

Launch anatomy

When each beat happens. Beats that weren’t found are left out; the full transcript isn’t published.

Launch anatomy for WillowRuntime 1:23. First five seconds: Other spoken line. Company named at 0s.0–5s hook · Other spoken lineCompany named at 0s0s Company named0:000:210:421:021:23
  • First five secondsOther spoken line
  • Company named0s

Not detected: product named, demo starts, call to action

Narration pace

155 words a minute · brisk

Narration pace: 155 words per minute; library median 142.SparseConversationalBriskDenselibrary median 142

Launches at this pace have a median of 327k views (n=215). Pace covers the whole runtime, pauses included.

Show the data
PaceMedian viewsVideos (n)
Sparse (0–80)571,51184
Conversational (80–140)346,258185
Brisk (140–180)326,715215
Dense (180–+)670,38071

How it compares

  • 0.7× the median benchmark or stat launch (1.2M, n=31)
  • 1.1× the median AI model launch (729k, n=48)
  • 2.1× the library median (407k, n=598)
Where this launch landsEach dot is one launch. The ringed dot is this one.
  1. All launchesn=598 · median 407k
  2. Benchmark or statn=31 · median 1.2M
  3. AI modeln=48 · median 729k

One launch Median This launchViews, log scale

Show the data
GroupVideosMedian25th pct75th pctMax
All launches598407,120103,0301,346,326138,357,050
Benchmark or stat311,233,026128,8492,273,08811,035,212
AI model48728,556326,7152,316,24539,909,098

Similar launches

All 31 benchmark or stat launches

Same hook style, closest view count, same category first.

  1. Freehand.aienterprise SaaSWe saved $260M in cash for cos. like Meta, Unilever, and J&J by finding baseless charges…Last year, Freehand recovered $260 million at Meta, Unilever, and J&J in overpaid invoices.Benchmark or stat1.2M views2:04Jul 29
  2. Gorgiasenterprise SaaSWe’ve raised over $100M to create AI customer support that doesn’t sound like AI slop…Introducing Gorgias AI Agent 3.0, the world's first customer support agent that outperforms human teams.Benchmark or stat1.1M views2:22Jul 21
  3. Pilot ProtocolAI agent1/ Today we're introducing Pilot Protocol: the Network OS for agents. A network where…250,000 users found Pilot on their own.Benchmark or stat655k views2:18Jul 27
  4. ElevenLabsAI modelToday we’re introducing Scribe v2: the most accurate transcription model ever released…Eleven Labs speech-to-text just had another major breakthrough with Scribe v2.Benchmark or stat558k views0:51Jan 9

Hook style, category, “why it worked” and the takeaway are AI-assisted labels, not hand-reviewed. Beat timings miss some moments: “not detected” does not mean it never happened.

Weekly, free

Get weekly insights on every new launch video

Every week, a breakdown of each new B2B launch video: how it hooks viewers in the first few seconds, and what worked.

Hook breakdown included

One short email a week.

Unsubscribe with one click, any time.

You are on the list. The first email arrives next week.

We use your email only to send this digest. See our privacy policy.

Videos belong to the companies that posted them and play through X’s official embed, which loads only after you click. Hook labels and “why it worked” notes are AI-assisted analysis and can be wrong. Curated from public launch posts on X. Data as of Nov 4, 2026. Corrections and removal