Gradium

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

Phonon, open-source on-device TTS models under 50M parameters

AI & MLProductionOctober 20, 2026@GradiumAI
0:00 · The Hook · Memory Address Hex Stream
0:00 / 0:00

Scene-by-scene timeline & spoken transcript

  1. The Hook

    Memory Address Hex Stream

    “(No spoken dialogue — low synth hum and digital ticking)”

    On screen
    7F389D 5498 9261 3558 ... 7F3F4D 7674 3467 7968
    Camera
    Static front-facing screen display
    Motion
    High-speed ASCII memory dump scrolling vertically
  2. The Hook

    Hardware Shelf CV Detection

    “(No spoken dialogue — synthetic scan chimes)”

    On screen
    [ 01 ] DETECTED [ 02 ] DETECTED [ 03 ] DETECTED [ 04 ] DETECTED [ 05 ] DETECTED
    Camera
    Static wide tabletop shot
    Motion
    Monochrome footage overlaid with target reticles and HUD tracking frames
  3. Problem Agitation

    Multi-Device Voice Synthesis Cascade

    “When in doubt, quantize. Houston, we have a tokenizer. Keep calm and run it locally. Cogito, ergo synthesize. Four score and seven milliseconds ago.”

    On screen
    [ 01 ] [ 02 ] [ 03 ] [ 04 ] [ 05 ] [ 06 ] [ 07 ] [ 08 ] [ 09 ] [ 10 ]
    Camera
    Static wide shot with sequential green highlight overlays
    Motion
    Selective color fill keyframing synchronized to audio voice prompts
  4. Product Reveal

    Unified Model Architecture Diagram

    “(No spoken dialogue — upbeat driving electronic beat)”

    On screen
    ONE MODEL ALL OF YOUR DEVICES
    Camera
    Wide architectural graphic view
    Motion
    Orthographic schematic grid with animated coordinate connector lines
  5. Feature Teaser

    On-Device Benchmark Showcase

    “The only thing we have to fear is a cache miss. I think, therefore I infer. A model divided against itself cannot batch efficiently. Less cloud, more compute. This sentence was generated locally.”

    On screen
    GALAXY S25 ULTRA 2025 SNAPDRAGON 8 ELITE BROWSER WEBASSEMBLY REACHY MINI 2025 ARM CPU MACBOOK PRO 2026 APPLE M5 JETSON ORIN NANO SUPER 2024 NVIDIA ARM + GPU ONEPLUS 5 2017 SNAPDRAGON 835 RUNS ON DEVICE
    Camera
    Rapid sequence of macro device shots
    Motion
    HUD callout cards with matrix LED dot waveforms and synchronized speech
  6. Feature Teaser

    Wireframe Tunnel Perspective

    “Great voices that run on every device. Available today.”

    On screen
    Great voices that run on every device Available today
    Camera
    Fast forward zoom through 3D vector wireframe tunnel
    Motion
    Tunnel wireframe lines receding into depth with floating device cards
  7. Call to Action

    Mosaic Collapse & Brand Identity

    “(No spoken dialogue — resolving low bass synth note)”

    On screen
    Phonon Gradium Gr
    Camera
    Center-locked graphic zoom
    Motion
    Collapsing video grid into letterform, sand/dust particle dispersion to brand mark

Related AI & ML Product Launches

Explore all AI & ML launches →
Tauvus
Hook 8.821.1M
TauvusAI & ML

Introducing Griffin, the first AI to pass the video Turing test

@tavus
Fish Audio
Hook 9.612.9M
Fish AudioAI & ML

The most emotive and controllable TTS model

@FishAudio
qvac
Hook 8.59.5M
qvacAI & ML

We just released QVAC TranscribePsy, a speech-to-text model small enough to run inside a phone app. It covers 25 European languages on a laptop and 8 on a phone, and on an iPhone 17 Pro Max's Ne

@qvac