Liva AI GTM brain

Liva AI supplies human voice and video datasets to teams developing realistic multimodal AI models.

category

Voice and video AI training data

market profile

  • text: Model builders need varied human speech and expressions that generic scraped datasets may not adequately represent.
  • text: Liva describes in-house collection of consented, rights-cleared voice and video for generative model training and fine-tuning.

personas

  • title: Head of speech or multimodal research
    pains:
    • Training data lacks diversity of accents, emotions and conversational contexts
    • A model sounds unnatural in real interactions despite strong benchmark scores
  • title: Training data or data operations lead
    pains:
    • Needs targeted human recordings at usable quality without building an entire collection operation
    • Must document consent, rights and dataset provenance for procurement
  • title: Founder of a voice AI product
    pains:
    • Needs more realistic conversational interactions for a specific application
    • Must assess whether custom data improves model performance enough to justify collection cost

icp

  • text: AI labs and startups building or fine-tuning speech, video or multimodal generation models, with a defined data gap and evaluation capability.
  • text: Companies integrating voice or video models and media production teams are additional stated audiences; size and geography are not specified.

message pillars

  • text: Lead with human voice and video training data.
  • text: Address coverage gaps in emotional and linguistic variation.
  • text: Emphasize real data and consent.

objections

  • objection: Can we use these recordings for our intended model and distribution?
    response: Review contributor consent, license scope, provenance and permitted use directly before procurement.
  • objection: Will the data actually improve our model?
    response: Agree target accents and use cases, evaluate a sample against a held-out baseline, then size a collection.
  • objection: Can you deliver enough of the exact scenarios we need?
    response: Confirm speaker coverage, recording specifications, quality controls and delivery milestones with the team.

competitors

  • name: Surge AI
    why: Supplies training data and human data services for AI development.
  • name: Hume AI
    why: Works on emotion-aware voice models and data, overlapping the expressive voice ecosystem.

influencers

  • name: Ashley Mo
    url: https://www.linkedin.com/in/ashley-mo
    why: Liva founder shares voice data and team updates.

keywords

  • voice AI training data
  • consented speech datasets
  • expressive speech dataset
  • multimodal video training data
  • voice model fine tuning data
  • human conversation audio dataset

voice

  • research led
  • human centered
  • direct

signals

  • text: A lab announces a speech or multimodal model and hires for data collection or model training.
  • text: A voice product identifies accent, emotion or conversational realism as a quality gap.

pricing

model: unknown

motion

sales_led

channels

  • YC launch
  • LinkedIn
  • research partnerships

confidence

medium

Founders

  • name: Ashley Mo
    title: Founder/CEO
    role: CEO
    bio: Co-founder & CEO | prev. Caltech CS | Published research on an AI model that could detect lung disease through cough recordings at 17 y/o (IEEE Xplore) and led large-scale research study to collect health audio data @MIT | Built biosensors for chronic kidney disease @Caltech
  • name: Aoi Otani
    title: Founder/CTO
    role: CTO
    bio: Co-founder & CTO

Back to YC companies