Grok Video 1.5
xAI's Grok Imagine video model — turn text, a first-frame image, or reference images into cinematic clips. Uniquely flexible duration: anything from 1 to 15 seconds.
Capabilities
| Feature | Support |
|---|---|
| Text-to-Video | Yes |
| Image-to-Video | Yes (first frame) |
| Reference-to-Video | Yes (up to 3 images) |
| Resolutions | 480p, 720p, 1080p |
| Aspect Ratios | Adaptive, 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3 |
| Duration | 1–15s continuous (1s granularity) |
| Frame Rate | 24 fps |
| Native Audio | Yes — always on (dialogue with lip-sync, ambience, SFX) |
| Seed | No |
| Negative Prompt | No |
The 1-Second Floor
Most video models bottom out at 3–4 seconds. Grok goes down to 1s, which unlocks:
- Micro-loops and stingers — logo reveals, UI transitions, social bumpers.
- Cheap motion tests — check whether a composition animates well before paying for a 10s take.
Adaptive Aspect
Like its image sibling, Grok Video supports an Adaptive ratio — the model infers the best frame from the prompt, or inherits the input image's shape in image-to-video mode.
Reference Inputs
Up to 3 reference images. In first-frame mode the image is a hard anchor (the clip starts exactly there); in reference mode images guide identity and style loosely.
Native Audio
Every clip ships with a synchronized soundtrack — there is no off switch. Dialogue is lip-synced to speaking characters, and ambience/SFX land on cue. Quote lines in your prompt to have characters speak them; describe the soundscape ("distant sirens, rain on glass") to shape the track.
Prompting Tips
- Lead with the camera. "Slow push-in", "orbit right", "crash zoom" — Grok follows camera grammar well.
- Write the sound too. Audio is always generated — an undirected soundtrack is a wasted channel. Quote dialogue, name the ambience.
- Match resolution to purpose. 480p for motion drafts, 1080p for delivery; the duration range is identical across tiers.
- Short clips want one beat. At 1–3s, prompt a single action ("the match ignites") — not a sequence.
Limitations
- Max 3 reference images; no video or audio reference inputs
- No first+last frame interpolation
- No seed / determinism
- No negative prompt
See Also
- Grok Image Pro — the image sibling (great for crafting the first frame)
- Seedance 2.0 — richer reference mixing and native audio