Grok App Update 1.3.54: Image & Video Upgrades, Speed Mode & Imagine 2.0
xAI has pushed a significant update to the Grok app — version 1.3.54 — delivering meaningful upgrades to Grok Imagine, the platform's image and video generation suite. The update brings sharper detail rendering, cinematic visual flair, and noticeably smoother motion to AI-generated video, alongside a faster overall app experience across chat, voice, and processing. It is the latest in a rapid series of improvements to a product that has become one of the fastest-growing AI creative tools of 2026.
Elon Musk has been vocal about the pace of development, noting that Grok models update approximately twice weekly — a cadence that puts most AI product roadmaps to shame. Alongside the version 1.3.54 rollout, Musk also confirmed that Imagine 2.0 is just a few weeks away, promising major upgrades in speech and audio processing and dramatically improved consistency in face and detail generation. And as of April 3, 2026, the first incremental improvement has already shipped: users can now choose between speed and quality modes when generating images, with a Professional mode coming later this month.
What version 1.3.54 actually changes
The visual quality improvements in this update are the headline. Grok Imagine's image generation now renders finer details with noticeably higher fidelity — users have described outputs like eyes reflecting the glow of city lights and cinematic lighting effects that previously required multiple regenerations to achieve. Video generation has been upgraded with smoother motion and more coherent frame-to-frame continuity, reducing the jitteriness that was a visible weakness in earlier versions.
The prompt examples circulating on X after the update capture the ambition of what the model can now handle: samurai battling dragons over neon skylines, macro-scale detail work on reflective surfaces, and stylized cinematic sequences that previously pushed the model into visible degradation. The improvements appear to be most pronounced in complex multi-element scenes where earlier versions struggled to maintain visual coherence.
Beyond the image and video quality gains, the update delivers meaningful performance improvements to the broader app experience. Chat response times are faster. Voice processing feels more responsive. The interface has been redesigned to surface video creation and image editing tools more prominently — a signal that xAI is positioning Grok Imagine as a front-and-centre feature rather than a secondary capability buried in the app.
The Aurora engine behind the visuals
The quality improvements in version 1.3.54 run on xAI's Aurora autoregressive engine — the same proprietary infrastructure that has powered Grok Imagine since its major 1.0 release in February 2026. Aurora was trained on a cluster of 110,000 NVIDIA GB200 GPUs, a compute investment that places it in the same hardware tier as the largest proprietary model training runs in the industry. The January 28, 2026 API launch made Aurora accessible to developers at $0.05 per second for 720p video with audio — roughly $0.50 for a 10-second clip — and the same engine powers the consumer-facing app improvements arriving today.
The Aurora engine's core differentiator is instruction-following accuracy. xAI has consistently benchmarked Grok Imagine against competitors on this dimension, and the latest update appears to push that edge further — particularly on complex compositional prompts where precise element placement, lighting direction, and motion style all need to be maintained simultaneously.
Speed, quality modes, and what's coming next
Alongside the visual quality improvements, April 3 also brought the first update Musk flagged as part of the Imagine 2.0 runway: users can now choose between speed and quality modes when generating images. Speed mode prioritizes fast output for rapid iteration. Quality mode dedicates more compute to each generation for higher-fidelity results. This is a practical addition for creators who iterate through many prompt variations before committing to a final output — speed mode lets them explore; quality mode lets them deliver.
A Professional mode is confirmed for later in April, suggesting a more advanced tier of generation capability is being staged in ahead of the full Imagine 2.0 rollout. Musk's framing for Imagine 2.0 — "a few weeks of training" rather than a fundamental rebuild — suggests the model architecture is largely set and the remaining work is a final training run focused on the two most requested improvements: face and detail consistency, and speech-audio synchronization.
Musk on Grok's accuracy edge
In the same communications around the update, Musk highlighted what he described as Grok's real-world accuracy advantage over competing models. His specific example: California zoning laws. Musk noted that when users ask Grok about California zoning regulations — a topic that requires integrating complex, frequently updated legal and municipal data — Grok correctly handles the nuance where other leading models have produced inaccurate or outdated responses. The example is pointed at a time when xAI is competing directly with OpenAI and Google for enterprise and professional users who need reliable, real-world factual accuracy rather than plausible-sounding answers.
Musk encouraged users to try Grok.com directly for chat, coding assistance, and real-time information — positioning the web interface as the primary destination for users who want to experience the full breadth of what Grok's models can do beyond the X social platform context.
How Grok Imagine has evolved in 2026
The pace of iteration on Grok Imagine since January 2026 has been one of the more notable product development stories in AI this year. The API launched January 28 with text-to-video, image-to-video, and video editing workflows. Version 1.0 shipped February 3, extending video length to 10 seconds at 720p with significantly improved audio. The Extend from Frame feature arrived March 2, letting users chain video clips by using the final frame of one generation as the starting point of the next — enabling continuous multi-clip sequences up to 15 seconds per segment. A folder organization feature shipped March 4. The version 1.3.54 quality push arrived this week. Imagine 2.0 is weeks out.
The usage numbers behind this iteration pace are significant. Grok Imagine generated 1.245 billion videos in January 2026 alone, and 314 million visits to the Imagine feature were recorded by early March. xAI's reach now spans approximately 600 million monthly active users across X and Grok apps. These are not hobbyist numbers — Grok Imagine has become a mainstream creative tool at a scale that justifies the compute investment and the rapid iteration cadence.
How to access the update
The version 1.3.54 update is available now on iOS and Android. Open the App Store or Google Play, search for Grok, and tap Update if it has not already installed automatically. The image and video generation improvements are available to X Premium subscribers. Basic X Premium starts at $8/month. SuperGrok and Premium+ subscribers get more daily generations and higher-quality output. The Grok Imagine API is accessible at api.x.ai for developers building applications, with pricing at $0.05 per second for 720p video with audio.
For the speed and quality mode toggle introduced today and the Professional mode arriving later in April, ensure you have the latest app version installed. The feature appears in the Grok Imagine interface as a mode selector before generation. Users on older app versions may not see the option until they update.