Nake
Settings
Language
ThemeAtelier
Privacy policyTerms of serviceRefund policy
Premium-50%50% off
Lite

Premium membership from $19 per month

Premium benefitsMore benefits
Fast generationHD outputAll image toolsAll video tools
Nake
  • Home
  • For You
  • AI Video
  • AI Image
  • Anime / Manga
  • Video Edit
  • Video Extend
  • AI Tools
  • Generate
  • Agent
  • Explore
  • Home
  • For You
  • AI Tools

Create & chat

  • AI Video
  • AI Image
  • Anime / Manga
  • AgentBeta
  • Explore
  • Generate
  • Video Edit
  • Video Extend+5s
Upgrade for more
    • Home
    • For You
    • Create
    • Pricing
    • More

    grok imagine video 1.5

    grok imagine video 1.5 is xAI's second-generation image-to-video model, built on the proprietary Aurora autoregressive engine. Released in May 2026, it took the #1 spot on the Image-to-Video Arena leaderboard with an Elo of approximately 1,330 — a +52 jump over version 1.0. The core mechanic is image-first generation: you supply a still frame — a photo, product render, concept art, or brand asset — write a motion prompt describing how that frame should move, and the model returns a short clip with native synchronized audio generated in the same inference pass. This page covers the model's documented capabilities, pricing, availability, and practical usage patterns based on xAI's official documentation and reported benchmarks.

    Official documentation
    Grok 2 logo
    Publisher identity: Grok 2 logo.

    On this page

    grok imagine video 1.5: Model Overview

    grok imagine video 1.5 is designed for short, audio-complete shots anchored to a strong still frame — not for long narrative sequences generated from text alone. It accepts image and video inputs but does not support text-to-video generation directly. The model outputs 480p or 720p resolution at 24fps, with clip lengths from 1 to 15 seconds. Native audio — sound effects, ambient sound, and music — is generated in the same inference pass, though it does not fire on every clip. The model is available via the xAI API in the us-east-1 region, with a rate limit of 10 requests per second (60 requests per minute).

    Architecture and Core Mechanic

    grok imagine video 1.5 is built on xAI's proprietary Aurora autoregressive engine. Unlike diffusion-based video models, Aurora generates frames sequentially, predicting each new frame based on the previous ones and the input conditions. This architecture is particularly well-suited for image-to-video tasks because it can maintain temporal consistency across the clip — subjects hold their appearance more reliably across all frames compared to version 1.0. The model accepts image inputs in JPG, PNG, and WebP formats, as well as video inputs. For best results, xAI recommends using a high-resolution source image with clean lighting and a clearly defined subject. The model does not support text-to-video generation; it requires a visual input to produce output. This design choice signals that this model is optimized for transforming visual input into premium video, not bootstrapping from scratch. The model's alias is grok-imagine-video-1.5-preview, with a versioned alias of grok-imagine-video-1.5-2026-05-30 reflecting the model state as of that date.

    Key Improvements Over Version 1.0

    Compared to the original Grok Imagine Video, version 1.5 delivers four documented gains. First, better facial accuracy — character identity holds across more frames, reducing the warping and distortion that plagued the earlier version. Second, tighter audio-visual sync — sound effects land on the action beat rather than near it, making the native audio feature genuinely useful for social content and drafts. Third, faster, more stable outputs with fewer artifacts in the opening seconds of generation. Fourth, wider stylistic range — surreal, illustrated, and photoreal sources are all handled confidently, whereas version 1.0 struggled with non-photoreal inputs. The combined effect of these improvements is the reported +52 Elo gain on the Image-to-Video Arena leaderboard, moving the model from mid-table to first place. The model also demonstrates stronger physics simulation for cloth, hair, water, and particle effects, and better adherence to complex, multi-clause prompts with several scene requirements.

    grok imagine video 1.5: Capabilities and Access

    grok imagine video 1.5 offers a focused set of capabilities centered on image-to-video generation with native audio. The model supports image and video inputs, video editing, and video extension. It does not support text-to-video generation. The model's strengths cluster around speed, cost efficiency, and output quality for short clips. Key capabilities include locking a cinematic grade onto motion, animating non-photoreal and illustrated art, triggering action physics from a single verb, and executing multi-part camera moves. The model also generates native audio — sound effects, ambient sound, and music — in the same inference pass, though this feature is not guaranteed on every clip.

    Image-to-Video Generation and Native Audio

    The primary capability of grok imagine video 1.5 is transforming a still image into a short video clip with synchronized audio. The model accepts a source image and a motion prompt describing how the frame should move. For example, a prompt like "Cinematic slow motion, dust particles swirl around the subject, dramatic backlighting, camera slowly pushes in, deep shadow contrast, volumetric haze" will produce a clip where lighting continuity holds across the entire duration and particulate physics read naturally. Native audio is generated in the same inference pass — sound effects, ambient sound, and music are timed to on-screen action without requiring a separate tool or manual sync. However, this feature does not fire on every clip. Reported testing puts synced SFX on the majority of clips but not all: action-heavy and locked-camera prompts return diegetic sound most reliably, while emotion-led narrative prompts more often return music only. For client-facing deliverables, a dedicated audio pass is recommended as a fallback.

    Camera Motion Control and Prompt Engineering

    grok imagine video 1.5 demonstrates sophisticated camera motion understanding compared to version 1.0. The model follows direct camera instructions reliably — prompts like "camera slowly orbits and pushes in" or "camera arcs right revealing the full facade" resolve accurately, while vague instructions like "dynamic shot" do not. Three prompt rules hold consistently across capabilities. Rule 1: Match prompt length to image strength — if the still already carries the look, use a short verb-led prompt; reserve long cinematic prompts for directing an aesthetic the source image doesn't already have. Rule 2: Name the camera move explicitly — the model follows direct camera instructions reliably. Rule 3: Use negatives to lock elements — "camera not moving" and "subject frozen" are more reliable than passive descriptions. The model also responds well to prompts that specify the how of the video, not just the what: explicit camera instructions, temporal anchors like "6 seconds" or "slow motion", and lighting descriptions like "golden hour" or "overcast diffuse light" produce more reliable results than broad color-based instructions alone.

    Availability

    grok imagine video 1.5 is available via the xAI API in the us-east-1 region. The model is in preview status, meaning the API interface and pricing are subject to change. It is accessible through the xAI developer platform with an API key. The model supports batch API in us-east-1 and us-west-2 regions. Rate limits are set at 10 requests per second (60 requests per minute). The model is also available through third-party platforms like JXP and AI/ML API, which offer their own pricing and credit systems.

    API Access and Regional Availability

    grok imagine video 1.5 is accessible via the xAI REST API in the us-east-1 region. Developers need an API key to make requests. The model ID is grok-imagine-video-1.5, with aliases grok-imagine-video-1.5-preview and grok-imagine-video-1.5-2026-05-30. The rate limit is 10 requests per second, which translates to 60 requests per minute. Batch API is supported in us-east-1 and us-west-2 regions, though batch pricing is listed as N/A in the current documentation. The model is also available through third-party platforms: JXP offers access with a credit-based system (2 credits per second at 480p, 3 credits per second at 720p), and AI/ML API provides a single key to access both Grok Imagine Video and this model Preview alongside 500+ other AI models. These third-party platforms may offer different pricing structures and rate limits than the direct xAI API.

    Pricing Structure

    grok imagine video 1.5 is priced at $0.08 per second of generated video at 480p resolution. At 720p, the price increases to $0.14 per second. At 1080p, the price is $0.25 per second, though the model's maximum output resolution is 720p — the 1080p pricing may apply to future capabilities or different model versions. Image input is priced at $0.01 per image. Audio input (preset voices) is free. Video input is priced per second. The model is available on multiple clusters with regional pricing variations. On third-party platforms like JXP, the cost is denominated in credits: 2 credits per second at 480p and 3 credits per second at 720p, with credit packs starting from $10 and no subscription required. Compared to the standard Grok Imagine Video model ($0.05/sec at 480p, $0.07/sec at 720p), the 1.5 Preview is approximately 60% more expensive at 480p and 100% more expensive at 720p. For content teams generating 50 clips per week, using the standard model for ideation and 1.5 Preview only for finals can cut overall video generation spend by 60-70%.

    Configurations

    This model offers two output resolutions — 480p and 720p at 24fps — with clip lengths from 1 to 15 seconds. The model accepts image inputs in JPG, PNG, and WebP formats, as well as video inputs. It does not support text-to-video generation. The model is available through the xAI API with a rate limit of 10 requests per second. Third-party platforms may offer different configurations and pricing structures.

    Output Resolution and Clip Length Options

    This model outputs video at either 480p or 720p resolution, both at 24 frames per second. There is no 1080p or 4K option in the current Preview release — this is a hard constraint that matters for high-end production. Clip length ranges from 1 to 15 seconds, with common lengths being 5, 10, or 15 seconds. The model does not support native clip chaining, so extended sequences require external stitching. For higher-resolution delivery, xAI recommends using Veo 3.1, which outputs true 4K and supports clips up to 140+ seconds. The 720p ceiling means this model is best suited for social content, rapid iteration, and animating strong still frames — not for final delivery at high resolution. The model's pricing varies by resolution: $0.08 per second at 480p and $0.14 per second at 720p. On third-party platforms like JXP, the cost is 2 credits per second at 480p and 3 credits per second at 720p.

    Input Formats and Supported Media Types

    This model accepts image inputs in JPG, PNG, and WebP formats. For best results, xAI recommends using a high-resolution source image with clean lighting and a clearly defined subject. The model also accepts video inputs, which are priced per second. Audio input (preset voices) is free. The model does not support text-to-video generation — it requires image or video as input to produce output. This is a key distinction from the standard Grok Imagine Video model, which supports text-to-video, image-to-video, video editing, video extension, and reference-to-video generation. In practice, the most effective workflow is to generate a base frame or short clip with the standard model (or with Grok's image generation capabilities), then pass that into this model for high-quality video output. The model's image input pricing is $0.01 per image, compared to $0.002 per image for the standard model. Video input is priced per second of input video.

    Practical Notes

    This model is a specialized tool for fast, audio-complete image-to-video generation at 480p and 720p. It excels at social content, rapid iteration, and animating strong still frames. However, it has clear limitations: a 720p maximum resolution, inconsistent native audio, facial softening under fast motion, and no native clip chaining. Understanding these constraints helps determine when to use this model versus alternatives like Veo 3.1, Sora 2, Seedance 2, or Wan 2.7.

    Best Use Cases and Workflow Integration

    This model is best suited for social content creators who need fast, audio-complete clips at 720p for TikTok, Reels, and Shorts. Brand and marketing teams can use it for rapid concept iteration — testing 10-20 visual ideas before committing to a 4K final. Designers and concept artists benefit from its ability to animate illustrated or surreal source art that other models degrade. Product teams can generate push-ins, rotations, and reveals from a single product render. The recommended two-stage production workflow uses the standard Grok Imagine Video model for concept generation (text-to-video, rapid iteration) and this model for final production (image-to-video, higher fidelity). This approach can cut overall video generation spend by 60-70% without compromising deliverable quality. For teams with on-premise requirements or high-volume budget constraints, Wan 2.7 is the practical alternative as it is the only open-source option, self-hostable under Apache 2.0 with no per-generation cost at scale.

    Limitations and Competitive Context

    This model has several documented limitations. The 720p maximum resolution is a hard constraint — no 1080p or 4K in the current Preview release. Facial softening occurs under fast motion: the faster the subject moves, the more facial detail degrades. Native audio is not guaranteed — it fires on the majority of clips but not all, with music-only outputs common on emotion-led and narrative prompts. The model caps at 15 seconds per generation, requiring external stitching for extended sequences. Preview behavior can shift as xAI may update output characteristics without notice. Compared to competitors: Veo 3.1 wins on resolution (4K vs 720p), audio quality (48kHz vs variable), and long-form capability (140+ seconds). Sora 2 is better for physics-heavy narrative shots. Seedance 2 supports multiple reference images and handles dialogue-driven content more reliably. Wan 2.7 is the only open-source option. The honest assessment: this model is a drafting and preview engine, not a final-delivery tool for high-resolution production. For everything that lives on a phone screen or in a pitch deck, it delivers more per credit than any competing model.

    FAQ

    Frequently asked questions about this model, covering capabilities, pricing, and practical usage.

    What should I know about grok imagine video 1.5?

    grok imagine video 1.5 is xAI's second-generation image-to-video model, built on the proprietary Aurora autoregressive engine. It was released in May 2026 and took the #1 spot on the Image-to-Video Arena leaderboard with an Elo of approximately 1,330. The model accepts image and video inputs and outputs 480p or 720p video at 24fps with clip lengths from 1 to 15 seconds. It generates native audio — sound effects, ambient sound, and music — in the same inference pass, though this feature does not fire on every clip. The model does not support text-to-video generation. Pricing is $0.08 per second at 480p and $0.14 per second at 720p. The model is available via the xAI API in the us-east-1 region with a rate limit of 10 requests per second. Key improvements over version 1.0 include better facial accuracy, tighter audio-visual sync, faster generation, and wider stylistic range. The 720p ceiling is the main limitation for high-end production. For final delivery at higher resolutions, Veo 3.1 is recommended.

    Original sources

    • Grok Imagine Video 1.5 | SpaceXSI Docs
    • Grok Imagine Video 1.5 Review: Is It Worth It in 2026?
    • Grok Imagine Video 1.5: Price, API, Specs & Data Policy | Cody
    • Grok Imagine Video vs Grok Imagine Video 1.5 Preview

    Related reading

    AI Video ToolsText-to-Image GenerationImage-to-Video Generation

    Read the sources for grok imagine video 1.5

    Creative tools & scenes
    • grok imagine nsfw
    • gpt image 2.0 nsfw
    • gemini nsfw
    • grok nsfw
    • hailuo nsfw
    • happyhorse nsfw