Official-source specification comparison · no winner declared
MiniMax H3 vs Grok Imagine 1.5
H3 documents multimodal video generation, while Grok Imagine 1.5 offers image-to-video plus separate fal text and image-reference routes. Start with the same source frame and motion brief for an image-to-video comparison; evaluate text-only and reference-set work separately rather than treating the original preview as Grok’s complete input scope. Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route
Evidence boundary: DualView has not yet run the controlled benchmark described below. Specifications come from official pages. Vendor samples are not matched for prompt, input, settings, selection, or post-processing. Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route
September 28 input-scope correction: The previous comparison relied on Grok’s image-to-video preview description. fal also documents explicit v1.5 text-to-video and image-reference routes. This is a documentation correction, not a new-release date or a measured quality result.
Official showcase material
These are real videos published by the model providers. Watch them as examples of each vendor's presentation, not as evidence that one model beats the other.
MiniMax H3 official showcase
MiniMax-published H3 output preserved without re-encoding.
Open the official sourceGrok Imagine 1.5 official showcase
xAI-published launch-page output preserved without re-encoding.
Open the official sourceWhat the official documentation actually supports
MiniMax positions H3 around unified context across text, images, video, and audio. xAI's 1.5 announcement is deliberately narrower: submit one still, describe motion, camera, pacing, atmosphere, physics, and sound design, then generate a continuation that stays faithful to the source. That preview description is historical: fal also documents v1.5 text and image-reference endpoints. Compare the same input task rather than assigning a quality advantage from a mode list. Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route
For a fair visual review, lock the input image, prompt text, duration, aspect ratio, and delivery resolution as closely as the APIs allow. Then inspect whether facial features, costume details, typography, lighting direction, and framing survive motion. Camera compliance and source-image fidelity should be scored separately because a model can preserve the frame while ignoring the requested movement, or move correctly while altering the subject.
The official pages differ in maturity and detail. MiniMax states H3 can generate up to 15 seconds at 2K with native stereo sound. The xAI launch article demonstrates a 10-second 720p request and says the preview supported clips up to 720p; current xAI model documentation lists 480p, 720p, and 1080p pricing. Treat those current docs as operational information, not as a retroactive launch claim. Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route
| Question | MiniMax H3 | Grok Imagine 1.5 |
|---|---|---|
| Fair overlapping lane | Image-conditioned video is part of a broader multimodal system | Image-to-video is one shared task; fal also documents text and image-reference routes |
| Other documented inputs | Text, image, video, and audio context | Starting image and prompt; separate text-only and image-reference endpoints on fal |
| Duration evidence | Up to 15 seconds | Official launch code example uses 10 seconds; the model page does not state a maximum |
| Resolution evidence | Up to 2K | Launch page: up to 720p; current model docs list pricing through 1080p |
| Audio direction | Native stereo sound | Prompt can describe sound design |
| Independent quality result | Not measured by DualView | Not measured by DualView |
Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route
Model profiles in one paragraph
MiniMax H3
MiniMax documents up to 15 seconds at 2K, native stereo sound, unified text-image-video-audio context, V2V motion transfer, and multimodal generation and editing. Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route
Grok Imagine 1.5
xAI’s preview announcement describes single-image animation. Additional fal v1.5 text and image-reference routes broaden the documented input choices; their schemas must be checked separately. Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route
Compare the endpoint as well as the model
As checked September 28, fal’s text schema lists 480p, 720p and 1080p. The reference schema lists up to seven images with 480p or 720p output. A limit from one route should not be copied to another. Its introductory audio-reference wording has no corresponding audio field in the displayed schema, so audio-reference submission remains unresolved here. Text schema · Reference schema
Keep still-image, text-only and reference-set comparisons separate. Save the exact endpoint, inputs and settings alongside each output, then inspect the matched files in DualView’s video editor. The vendor showcases above do not constitute those matched experiments.
A fair comparison protocol
The protocol below isolates shared capabilities before exploring model-specific advantages. Save original files and settings so another reviewer can reproduce the result. Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route
Use portrait, product, typography, and wide-environment source frames with licensing recorded. Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route
Ask for the same push-in, orbit, subject action, atmosphere, and synchronized sound event. Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route
Compare at the same resolution first; keep H3's 2K test as a separate delivery-resolution lane. Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route
Score identity retention frame by frame, motion adherence, unwanted reframing, audio fit, and artifacts. Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route
Which one belongs on the shortlist?
Use a matched image-to-video task when production starts from an approved frame; Grok’s separate text and image-reference routes are additional candidates for their respective tasks. Test H3 when the shot depends on mixed references, V2V motion, native stereo, or 2K delivery. For shared I2V work, neither documentation set replaces a matched output review. Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route
What this page does not claim
This MiniMax H3 and Grok Imagine 1.5 review does not declare an overall quality winner or turn provider preference figures into independent evidence. It leaves undocumented limits blank. Recheck the linked first-party pages because access, aliases, pricing, and output options can change. Provider announcement · xAI preview · xAI current docs · fal text route · fal reference route