xAI Grok Imagine Video 1.5 Modes — Pricing Impact

xAI has expanded Grok Imagine Video 1.5 into a three-mode video API. The release note, published under xAI’s July updates and detected on August 1, confirms text-to-video, image-to-video, and reference-to-video, plus native Full HD output for two of those workflows.

The change makes version 1.5 more useful for production creative pipelines, but it is not a free upgrade. The live table above shows a higher per-second rate than the previous model at every shared resolution. Teams should compare usable clips per dollar, not generation price alone.

What changed

Text-to-video turns a prompt into a new clip. Image-to-video animates a supplied first frame. Reference-to-video uses one or more images to guide people, objects, clothing, or visual identity without forcing the reference to become the opening frame.

Both text-to-video and image-to-video now support native 1080p. Reference-to-video tops out at 720p. All three modes accept clips from 1 to 15 seconds, configurable aspect ratios, and asynchronous generation through the Videos API.

Reference-to-video also accepts up to three built-in xAI voices. Developers tag them in the prompt as <AUDIO_0> through <AUDIO_2>. This is preset-voice conditioning, not voice cloning: uploaded audio is not accepted, and xAI currently limits the feature to trusted partners in the United States.

Pricing impact

The generated comparison above reads current resolution rates from our canonical pricing dataset. At the base 480p tier, version 1.5 carries a 60% premium over the previous model. At 720p, the premium doubles the per-second rate. 1080p is a new 1.5-only tier, so there is no like-for-like price on the older model.

xAI’s simplified Imagine pricing page displays the base rate, while its public model configuration exposes resolution-specific billing. Budget with the resolution row you actually request. A 10-second job costs ten times that row before any applicable image-input charge, retries, discarded generations, storage, or post-production.

The xAI Batch API does not lower this bill: xAI says its batch discount applies to text and language models only. Video jobs can be submitted in batches, but they retain standard video rates.

Who benefits—and who pays more

Creative-tool builders benefit from one API covering storyboards, still-image animation, product placement, virtual try-on, and character-consistent clips. Native Full HD removes a separate upscale step for text- and image-led workflows. Teams already using xAI’s Files API can reuse stored images by file_id, which can simplify repeatable campaigns.

The cost increase lands on high-volume social, advertising, and catalog teams, especially when they move existing 720p workloads to version 1.5. Reference-to-video users also need to accept the lower resolution ceiling, even though text- and image-to-video can reach Full HD.

This announcement includes no quality, prompt-adherence, temporal-consistency, or generation-speed benchmark. The new modes and resolution are verified capabilities; they do not prove a lower cost per approved clip.

What developers should do now

1. Pin the exact model ID instead of assuming an unversioned alias will preserve behavior or billing.
2. Test the same prompt set at 480p, 720p, and 1080p; record approval rate, retries, latency, and final delivery resolution.
3. Use 720p for reference-to-video tests because that mode cannot produce native Full HD.
4. Cap duration during iteration, then render the approved prompt at the final length and resolution.
5. Save completed output promptly: xAI-hosted video URLs are temporary unless you configure your own storage path.
6. Read cost_in_usd_ticks from every completed response and alert on unexpected spend.

For the wider model catalog, see the live xAI Grok pricing page. The token calculator is useful for the text models surrounding a video workflow, but video generation must be budgeted per second. Compare the broader market in our AI API pricing guide and review OpenAI API pricing for the reasoning or scripting layer.

Voice alternative

If your workflow needs a separate voiceover, dubbing, or multilingual audio layer instead of xAI’s limited preset voices, test ElevenLabs on the same finished clips and compare total production cost.

Affiliate disclosure: we may earn a commission from the sponsored link above. It does not affect the pricing analysis.

Bottom line

Grok Imagine Video 1.5 is now a broader production API, not merely a text-to-video endpoint. Native 1080p and reference-guided generation expand what teams can ship, while higher resolution rates make disciplined testing essential. Start with short, lower-resolution candidates and promote only approved prompts to the final render.

Sources: xAI’s official release notes, video generation guide, image-to-video guide, reference-to-video guide, and Imagine pricing.