Launch Directory/AI & ML/Startup Launch
Startup Launch

Startup Launch · Launch Video Breakdown: Hook, Pacing & Motion Design

Today we’re releasing OCRia-1, a frontier VLM for document parsing that sets a new SOTA in accuracy, cost, and latency. It not only sets a new high score on ParseBench (87.4), but also leads in thre

0:00 · The Hook · Introducing OCRia-1
0:00 / 0:00
AI & MLProductionOctober 22, 2026

Founder Direct-to-Camera Announcement

15.8

Cuts / min

0:53

Runtime

58K

Views

804

Likes

0

Reposts

0

Replies

Scene-by-scene timeline & spoken transcript

  1. The Hook

    Introducing OCRia-1

    “Today we're introducing OCRia-1, our state-of-the-art model for document parsing.”

    On screen
    INTRODUCING OCRia - 1 SOTA model for document parsing
    Camera
    Static two-shot cutting to full-screen 2D motion graphic title card
    Motion
    Pixel-art dot matrix grid reveal with typewriter text
  2. Problem Agitation

    Built for Tables & Cria Etymology

    “Built for the hardest part of any document: the tables. Like a cria, it holds up on its own for even the hardest conditions.”

    On screen
    Built for tables Cria /'kri:.ə/ • A cria is a baby llama. Wikipedia +1 • OCR + cria = OCRia-1 Jerry Liu Co-Founder and CEO, LlamaIndex
    Camera
    Medium two-shot cut to document fan-out cut to medium close-up on Jerry
    Motion
    Document scatter layout followed by dictionary-style tooltip fly-in
  3. Problem Agitation

    Enterprise Table Complexity

    “So much of what businesses depend on live in tables: balance sheets, insurance schedules, financial filings. After years of parsing billions of pages in LlamaParse, we've seen just about every kind there is.”

    On screen
    Simon Suo Co-Founder and CTO, LlamaIndex Billions of pages Every kind there is
    Camera
    Medium close-up on Simon with cascading financial document stack
    Motion
    Stacked card cascade and pixel dissolve transition
  4. Product Reveal

    Training Dataset Scale

    “So we trained OCRia-1 on the hardest tables we could find, and graded it on structure and meaning, not just text.”

    On screen
    84K+ PAGES 11K+ TABLES 64K+ TABLES 10K+ CHARTS 100K+ EQUATIONS 20+ LANGUAGES
    Camera
    Medium close-up on Jerry with graphic overlay on left
    Motion
    Incremental metric counter pop-ins beside pixel mascot
  5. Feature Teaser

    Benchmark Dominance

    “OCRia-1 tops every table benchmark we tested, ours and third party included. It is the best model for parsing documents with complex tables.”

    On screen
    PARSEBENCH 1st +12.1 OCRia-1 4B 83.6 FR-BENCH-PDF2MD 72.1 ParseBench Tables • Score x 100 91.5 olMOCR-bench 92.9 fr-bench-pdf2md PulseBench-Tab 84.1
    Camera
    Close-up cut to series of overlapping benchmark data cards
    Motion
    Staggered sliding cards with animated horizontal bar charts
  6. Feature Teaser

    LlamaParse Availability & Demo

    “And it is live in LlamaParse today, pushing a new frontier in price performance.”

    On screen
    OCRia-1 Available today
    Camera
    Two-shot cutting to animated financial table scan
    Motion
    Pixel llama scanning across table cells highlighting extracted data
  7. Call to Action

    Enterprise Credit Giveaway & Countdown

    “So for the next seven days we're giving away up to $3,000 in credits for enterprises. Bring us your worst docs. Clock's running.”

    On screen
    $3,000 Bring us your worst docs $3,000 in credits LlamaIndex $3,000 in credits 48:00:00 llamaindex.ai/contact
    Camera
    Two-shot of founders standing and walking off, revealing end card
    Motion
    Typewriter numeric overlays transitioning to live countdown timer

Related AI & ML Product Launches

Explore all AI & ML launches →
Grok Bot 1.0 — AI teammates for your team
@bot
AI & ML
@bot
Grok Bot 1.0 — AI teammates for your team@bot
Tauvus
@tavus
AI & ML
@tavus
Tauvus@tavus
Fish Audio
@FishAudio
AI & ML
@FishAudio
Fish Audio@FishAudio