Adobe Inc. and Speechmatics, a provider of speech recognition technology, announced a significant expansion of their partnership on April 21, 2026, with the release of a new on-device speech-to-text (STT) model for Adobe Premiere Pro. This integration marks a shift in video editing workflows by providing users with transcription capabilities that match the accuracy of cloud-based engines without requiring an internet connection or external data transfer.
The new STT engine is designed to address the stringent security and privacy requirements of large-scale enterprises and government agencies. By processing audio data locally on the user’s hardware, the system ensures that sensitive intellectual property and confidential recordings remain within the organization’s firewall. This development follows increasing demand from the media and entertainment sector for tools that comply with global data protection regulations, such as GDPR and CCPA, while maintaining high-speed production cycles.
According to technical specifications released by the companies, the on-device model utilizes a compressed version of Speechmatics’ Autonomous Speech Recognition (ASR) technology. Despite the reduced footprint required for local execution, the model achieves word error rates (WER) comparable to standard cloud-grade services. The integration supports over 30 languages and includes features such as speaker identification and automated punctuation. This allows for seamless captioning and indexing of video content directly within the editing timeline.
Deepa Subramaniam, Adobe’s Vice President of Product Marketing for Creative Cloud, stated that the collaboration reflects the company’s commitment to providing professional editors with tools that balance performance and security. Subramaniam noted that the ability to generate accurate transcripts and captions instantly, regardless of connectivity, is a critical requirement for modern broadcast and film production environments where data sovereignty is a primary concern.
Speechmatics CEO Cantillon emphasized the engineering milestone of bringing large-scale neural networks to edge devices. The partnership leverages modern hardware acceleration found in the latest workstation processors and dedicated AI chips. This allows Premiere Pro to handle transcription tasks in the background without significantly impacting the performance of other editing functions, such as color grading or rendering. The companies reported that the model was optimized specifically for the silicon architectures common in professional creative workstations.
The update is being rolled out to Adobe Creative Cloud enterprise subscribers starting today. It follows a multi-year collaboration between the two firms, which previously focused on cloud-integrated transcription services. Adobe confirmed that the on-device feature will be available as a standard component of the Premiere Pro toolset, requiring no additional licensing fees for existing enterprise-tier customers. This move aligns with Adobe’s broader strategy to integrate advanced AI capabilities directly into its creative suite while mitigating the risks associated with third-party data processing.