GPT-56T 861 —
MUSE-SPK 837 —
GPT-56SC 789 -0.1%
GLM-5 781 —
CL-OP55X 779 -0.1%
GROK-46H 779 -0.1%
QWEN-38X 748 —
GPT-6A 743 —
KIMI-K3X 742 —
CL-FAB5H 697 -0.1%
CL-OP5H 674 -0.1%
GEM-38FH 672 —
CL-OP5X 669 -0.1%
CL-OP55H 667 -0.1%
CL-OP46H 656 -0.2%
CL-OP47H 647 -0.2%
GPT-56S 617 -0.2%
GEM-37FH 609 -0.2%
GEM-36FH 592 -0.2%
CL-OP48H 587 -0.2%
CL-OP47 580 -0.2%
GEM-35FH 579 -0.2%
GPT-55H 540 -0.2%
INKL 531 —
GEM-31P 511 -0.2%
CL-OP46 498 —
GEM-3P 498 —
CL-OP48 492 —
GPT-52 464 —
GPT-55 423 —
GPT-56T 861 —
MUSE-SPK 837 —
GPT-56SC 789 -0.1%
GLM-5 781 —
CL-OP55X 779 -0.1%
GROK-46H 779 -0.1%
QWEN-38X 748 —
GPT-6A 743 —
KIMI-K3X 742 —
CL-FAB5H 697 -0.1%
CL-OP5H 674 -0.1%
GEM-38FH 672 —
CL-OP5X 669 -0.1%
CL-OP55H 667 -0.1%
CL-OP46H 656 -0.2%
CL-OP47H 647 -0.2%
GPT-56S 617 -0.2%
GEM-37FH 609 -0.2%
GEM-36FH 592 -0.2%
CL-OP48H 587 -0.2%
CL-OP47 580 -0.2%
GEM-35FH 579 -0.2%
GPT-55H 540 -0.2%
INKL 531 —
GEM-31P 511 -0.2%
CL-OP46 498 —
GEM-3P 498 —
CL-OP48 492 —
GPT-52 464 —
GPT-55 423 —
← Back to feed

xAI Opens Grok Voice APIs: $0.10/hr Transcription, $4.20/M Characters Speech Synthesis

xAI shipped two standalone audio APIs on April 18 — Grok Speech-to-Text (STT) and Grok Text-to-Speech (TTS) — both available immediately through the standard xAI developer console. The infrastructure is not new: it has been running under load at scale since Grok Voice launched on mobile, across Tesla vehicle fleets, and inside Starlink’s customer support systems. What’s new is that developers can now call it directly.

Grok STT: The Numbers

The Speech-to-Text API covers transcription across 25 languages with both batch and streaming modes. Pricing is $0.10 per hour for batch and $0.20 per hour for streaming. The API accepts 12 audio formats up to 500 MB per request and ships word-level timestamps, speaker diarization, and multichannel support out of the box.

The headline benchmark is phone call entity recognition, where xAI reports a 5.0% error rate — a result the company says beats ElevenLabs. That’s a specific, verifiable claim on a domain that matters most to call-center and voice-agent developers, the exact segment currently held by Deepgram and AssemblyAI.

Intelligent Inverse Text Normalization — correctly rendering numbers, dates, currencies, and addresses — is included natively. This feature routinely shows up as a friction point in transcription pipelines that don’t handle it well.

Grok TTS: The Numbers

The Text-to-Speech API is priced at $4.20 per million characters. Five expressive voices are available: Ara, Eve, Leo, Rex, and Sal, with Eve set as the default. The API supports 20 languages with a 15,000-character limit per REST request; longer content runs over WebSocket streaming.

Speech tags — inline modifiers embedded in the text — give developers granular control over delivery style, pace, and tone without needing a separate audio editing pass. The approach mirrors what Google shipped with Gemini 3.1 Flash TTS three days earlier, though xAI’s is built on a narrower voice roster with a focus on low-latency production use.

Why This Matters

The speech API market is one of the few AI infrastructure categories where no single provider has locked in dominance. ElevenLabs holds mindshare on quality; Deepgram and AssemblyAI hold enterprise STT accounts; OpenAI’s Whisper-based endpoints dominate DIY pipelines. xAI enters with a credible production pedigree — real usage under real load — rather than a research release.

The pricing is aggressive. $0.10/hr batch for STT undercuts Deepgram’s comparable tier. $4.20/M characters for TTS is competitive with ElevenLabs, which charges more for its highest-quality voices.

Grok 4 and later models are the obvious future upsell: developers who build voice pipelines on the xAI STT/TTS layer are a natural funnel toward the full Grok API stack. xAI is building a surface area, not just shipping a product.