SpaceXAI released a new speech-to-text API.
Grok Voice Transcribe 2.0 arrived on September 18, 2026, with a claim that will attract attention: twice the accuracy of Grok Voice Transcribe 1.0 at the same price. The release also adds new features, though the available details do not specify what those features include.
SpaceXAI describes Grok Voice Transcribe 2.0 as an improved speech-to-text service. Its central promise is straightforward—more accurate transcription without a higher price for the comparison with version 1.0.
“Grok Voice Transcribe 2.0 claims 2x the accuracy of 1.0 at unchanged pricing.” That is the headline proposition, and the wording matters: this is an accuracy claim, not a supplied measurement across named languages, audio conditions, or evaluation tests.
Pricing splits batch and streaming transcription
Grok Voice Transcribe 2.0 costs $0.10 per hour for batch transcription and $0.20 per hour for streaming. The two prices separate work completed in batches from transcription delivered as audio streams move through the service.
That pricing gives the release a clear structure without pretending every transcription task is identical. Batch transcription costs less, while streaming costs twice as much per hour—an uncomplicated distinction in an industry that often manages to make billing feel like a side quest.
SpaceXAI also says the API is built on the same foundation behind its voice product. “Grok Voice Transcribe 2.0 is built on the audio foundation model behind Grok Voice.” The statement connects the transcription API to Grok Voice, but it does not provide additional details about the model or its evaluation.
Microsoft and Atlassian Loom extend its reach
The release is not limited to SpaceXAI’s own services. Atlassian Loom uses Grok Voice Transcribe 2.0 for video transcription, giving the API a stated use in recorded video.
Microsoft has also integrated Grok into Word, Excel, and PowerPoint as a subprocessor. That places Grok inside three Microsoft products, although the supplied facts do not describe the exact features, workflows, or user experiences involved.
The Microsoft integration matters because it changes the shape of the announcement. Grok Voice Transcribe 2.0 is presented as an API, but Grok also appears in Word, Excel, and PowerPoint through Microsoft’s subprocessor arrangement. Atlassian Loom adds video transcription to that picture, creating a short list of concrete integrations rather than another announcement built only from pricing and ambition.
Still, the strongest performance claim remains unverified within the available details. SpaceXAI says version 2.0 is twice as accurate as version 1.0 at unchanged pricing, but the release provides no additional accuracy figures beyond that comparison.
The economics are easier to read. Batch transcription costs $0.10 per hour, streaming costs $0.20 per hour, and the accuracy claim says version 2.0 improves on version 1.0 without changing its price. Those figures give potential users a simple starting point for comparing the two transcription modes.
For now, the release rests on three facts: SpaceXAI launched Grok Voice Transcribe 2.0 on September 18, 2026; the API claims twice the accuracy of version 1.0 at the same price; and the service costs $0.10 per hour for batch transcription or $0.20 per hour for streaming.
The integrations add weight, but they do not answer every question. Atlassian Loom uses the service for video transcription, while Microsoft has integrated Grok into Word, Excel, and PowerPoint as a subprocessor. The numbers are clear. The proof behind the accuracy claim is not.
Based on
