Grok Voice is live on fal. The latest speech model from @SpaceXAI that answers in 0.70 seconds and finishes its tool calls before the sentence ends. Transcription with word-level timestamps, text to speech in 30+ voices across 25+ languages, and cloning from two minutes of audio. Every voice in this video is from Grok Voice.

Sep 16, 2026 · 7:10 PM UTC

34
26
252
210,176
Join fal and @SpaceXAI on September 22 for a live workshop on building your first voice agent with Grok. Sign up: luma.com/spacexai-l6cm

SpaceXAI x Fal: Build Your First Voice Agent with Grok · Luma

Join SpaceXAI and Fal for a hands-on workshop where you’ll build real-time voice agents with the same models running in production at Starlink, Tesla, Gopuff,…

luma.com
1
2
35
7,034
Try Grok Voice on fal. Speech to speech: fal.ai/models/xai/grok-voice Real time: fal.ai/models/xai/grok-voice…

Grok Voice (Speech to Speech) API on fal

Build real-time voice applications powered by Grok. Send recorded audio and get a voice and text response.

fal.ai
1
12
2,055
Sort replies: Relevant Recent Liked
Replying to @fal @SpaceXAI
Good team work! @SpaceXAI x @fal
3
377
Replying to @fal @SpaceXAI
0.70s latency is the crazy part
321
Replying to @fal @SpaceXAI
Exciting times 🦾 happy to see what’s happening
348
Replying to @fal @SpaceXAI
0.7 second response time, that's fast
166
Replying to @fal @SpaceXAI
tool calls finishing mid-sentence is wild
1
258
Replying to @fal @SpaceXAI
👀
43
Replying to @fal @SpaceXAI
Check ur dms fal
225
Replying to @fal @SpaceXAI
It works awesome on our platform. Thanks for nailing the human latency. It responds in similar fashion and natural speed as a human with almost-perfect intonation and little to no lag. We even tested it on a slower 10mbps connection and it was still passable. Crushed it
8
Replying to @fal @SpaceXAI
What is the voice cloning endpoint? I don’t see it anywhere.
51
Replying to @fal @SpaceXAI
Wow this is amazing
63
Replying to @fal @SpaceXAI
Works great as our customer service. Wiring it into onboarding next.
69
Replying to @fal @SpaceXAI
Not the best
37
Replying to @fal @SpaceXAI
Grok Voice answering in 0.70 seconds with tool calls before the sentence ends is absolutely next level
63
Replying to @fal @SpaceXAI
@grok is this a fee added to the fee you pay to use grok voice 2.0 api?
1
126
Replying to @fal @SpaceXAI
word level timestamps are the useful part here. on launch films the voice was never the problem, landing the cut on the right syllable was. do the timestamps stay stable if you regenerate the same line, or do the boundaries move every run?
45
Replying to @fal @SpaceXAI
this could genuinely cut down response times for customer support a lot. voice agents that are actually fast enough to feel natural are still rare
31
Replying to @fal @SpaceXAI
The latency claim is compelling because it changes the interaction model, not just the benchmark. Word-level timestamps also make partial-result UX and interruption handling practical from day one.
282