.Wave GTM brain

.Wave provides dedicated inference capacity and streaming speech APIs for teams building real-time AI applications.

category

Real-time AI inference infrastructure

market profile

  • text: Live voice workloads need predictable latency and concurrency, while fixed GPU reservations may sit idle off peak.
  • text: .Wave offers both scheduled dedicated capacity and serverless speech models, with the latter in public beta.

personas

  • title: Voice AI engineering lead
    pains:
    • Long or inconsistent response times interrupt conversational flow.
    • Concurrent streams strain GPU serving infrastructure.
  • title: AI infrastructure or platform lead
    pains:
    • Paying for peak GPU capacity around the clock wastes budget during quiet hours.
    • Needs predictable latency, regional placement and a capacity agreement.
  • title: Developer building voice agents
    pains:
    • Must integrate speech recognition and live audio quickly without running GPUs.
    • Needs to understand session limits and fallback behavior before production.

icp

  • text: Software companies building voice agents, live transcription or conversational products with latency-sensitive workloads.
  • text: Teams with predictable peaks and a need for reserved concurrent calls or requests per second; no company size minimum stated.

message pillars

  • text: Make response-time commitments for dedicated inference.
  • text: Align capacity reservations to the workload.
  • text: Offer hosted models for live speech applications.

objections

  • objection: What happens if the beta reaches capacity?
    response: Test realistic loads and keep a fallback for critical workflows; the documentation warns that beta sessions can be refused.
  • objection: How do we know dedicated capacity saves money?
    response: Compare a tailored reservation against peak and off-peak traffic, latency commitments and existing GPU spend.

competitors

  • name: Deepgram
    why: Provides streaming speech recognition APIs and is named as a migration alternative in .Wave documentation.
  • name: Pipeshift
    why: Provides inference infrastructure competing for model hosting and latency-sensitive deployments.

keywords

  • real time speech inference
  • streaming speech to text API
  • full duplex voice API
  • dedicated GPU inference latency SLA
  • voice agent concurrency
  • speech recognition per minute pricing

voice

  • Technical and concrete
  • Direct, performance-oriented

signals

  • text: Voice agent product is moving from a prototype to production and needs more concurrent sessions.
  • text: GPU capacity is overprovisioned outside predictable peak hours.

stack

  • WebSocket
  • Python
  • TypeScript
  • NVIDIA Nemotron

pricing

model: usage
starting price: $0.00045 per minute for streaming ASR; $10 free credits. Dedicated capacity is quoted.

motion

hybrid

channels

  • developer documentation
  • LinkedIn
  • X

confidence

high

Founders

  • name: Thomas Minassian
    title: Founder/CEO
    role: CEO
    bio: CEO .Wave. Our team built Parler-TTS Mini Multilingual, the world’s first fully multilingual open-source text-to-speech model, with 300,000 monthly downloads, and Dollyglot, the first model to turn a photo into a real-time video avatar.
  • name: Paul-Henri Biojout
    title: Founder/CTO
    role: CTO
    bio: CTO .Wave. Our team built Parler-TTS Mini Multilingual, the world’s first fully multilingual open-source text-to-speech model, with 300,000 monthly downloads, and Dollyglot, the first model to turn a photo into a real-time video avatar.

Back to YC companies