Resources // Voice & Multimodal AI

Deepgram Startup Program

  • High-value
Offer $100,000 credits
Suits for Non-VC-backedVC-backed
Updated Jun 2026
Apply ↗

The offer

  • Voice AI APIs: Speech-to-Text, Text-to-Speech, Voice Agent, Audio Intelligence (callbots, voice UIs, transcription, analytics)
  • Up to $100,000 in credits over 12 months (credits must be used within 12 months)
  • Community and networking (invite-only events, engineering support groups, Discord), plus access to new releases
  • Public free tier anyone can use: $200 in free credits, no card (~12,000 minutes). AWS Activate also lists the $100k offer.

Who qualifies

  • An AI builder or early-stage startup, in production or launching within ~6 months, committed to engaging with the community
  • Apply with your Deepgram Project ID; review-based (the team typically responds within a week)
  • Soft funding ceiling around <$10M / pre-Series A, and a lean toward new Deepgram customers (per the application FAQ)

Community Insights

Deepgram is the low-latency voice-agent default. Its Nova/Flux STT models top independent WER benchmarks and stream fast enough that text appears as you speak, and the $200 public free tier (~12k minutes) plus the $100k startup credit make it easy to start. The honest trade-offs: it is tuned for real-time more than maximum offline accuracy (some pair it with Whisper/AssemblyAI for batch re-transcription), batch transcription can be far pricier than self-hosting, and the jump from free to the Growth tier is steep.

Best Practices (from community tips)

  • Use Deepgram for real-time/voice-agent work where latency matters; consider re-transcribing recordings with a cheaper/accuracy-first engine for analytics.
  • Burn the $200 free tier first to validate accuracy on your actual audio (accents, jargon, noise) before committing the startup credit.
  • Apply with your Project ID and note the soft under-$10M / new-customer lean; ask about AWS Activate for additional credits.
  • Model batch cost vs self-hosting (Whisper/NeMo). For high-volume offline transcription, Deepgram can be 10x or more pricier.

Community Reviews

Positive

Negative

Neutral / Mixed