Grok Imagine Video: classic vs 1.5, modes, pricing
Grok Imagine Video is xAI’s video model family and Playcut’s budget video tier. Two versions are available: classic (grok-imagine-video, 20 credits/second) and 1.5 (grok-imagine-video-1.5-preview, 32 credits/second). Both cap out at 720p. Pick classic if you need reference-to-video or clip extension — it is the only Grok version that supports them. Pick 1.5 if you are animating a still image and want the better result; it is image-to-video only — it hard-rejects both reference images and text-to-video.
Specs at a glance
Section titled “Specs at a glance”| Imagine Video (classic) | Imagine Video 1.5 | |
|---|---|---|
| Model id | grok-imagine-video | grok-imagine-video-1.5-preview |
| Provider | xAI | xAI |
| Modes | t2v, i2v, r2v, extension | i2v only (t2v verified rejected by xAI, 2026-07-21) |
| Aspect ratios | 16:9, 4:3, 1:1, 9:16, 3:4, 3:2, 2:3 | same 7 offered, unverified on 1.5 |
| Resolutions | 480p, 720p | 480p, 720p |
| Duration | t2v 1–15 s · i2v / r2v / extension 1–10 s | i2v 1–10 s |
| Default duration | 6 s | 6 s |
| Max reference images | 7 | 0 |
| Credits | 20 / second | 32 / second |
No 1080p and no 4K exist anywhere in the Grok video family. If you need those, use Veo 3.1 or Seedance 2.0.
Which version should I use?
Section titled “Which version should I use?”| If you want to… | Use |
|---|---|
| Animate a still image at the best quality Grok offers | 1.5 |
| Drive a video from up to 7 reference images | classic — 1.5 cannot do it at all |
| Extend an existing clip | classic — the only Grok version wired for extension |
| Generate from text at the lowest price per second | classic (20 cr/s vs 32 cr/s) |
| Use a non-16:9 aspect ratio with confidence | classic — 1.5’s ratio list is inherited, not verified |
1.5 costs 1.6× classic per second. Classic is the cheapest video model in the Playcut catalogue.
Text to video (classic only)
Section titled “Text to video (classic only)”Prompt only. Duration 1–15 seconds, resolution 480p or 720p.
1.5 does not support this mode: a live probe (2026-07-21) confirmed xAI rejects it with “Text-to-video is not supported for this model,” so Playcut blocks 1.5 text-to-video before any credits are held. Use classic or Veo for text-to-video.
Image to video (both versions)
Section titled “Image to video (both versions)”One source image plus a prompt. The image pins the first frame of the clip — this is a frame lock, not a style hint. Duration is capped at 10 seconds on this mode.
This is 1.5’s headline strength and the main reason to pay the extra 12 credits per second.
Reference to video (classic only)
Section titled “Reference to video (classic only)”Up to 7 reference images plus a prompt. References guide appearance and style without locking any frame. In your prompt you can address them positionally as <IMAGE_1>, <IMAGE_2>, and so on. Duration is capped at 10 seconds.
xAI publishes no documented upper bound on reference count; 7 is a Playcut-imposed ceiling, and the API rejects an eighth.
Video extension (classic only)
Section titled “Video extension (classic only)”One existing video asset plus a prompt. durationSeconds here is the length of the extension only, not of the finished clip — a 10-second input with durationSeconds: 5 produces a 15-second output. You are billed for the added seconds, not for the total. Extension length is capped at 10 seconds per call server-side — but note that the public generate-video-extension tool schema currently locks durationSeconds to 8, so over the API an extension is always 8 seconds.
Not available on either version
Section titled “Not available on either version”- Interpolation — Veo 3.1 only on Playcut.
- Video edit — Wan 2.7 only on Playcut. The Playcut API returns a 400 for Grok video edit even though xAI ships an edit endpoint.
Pricing
Section titled “Pricing”Playcut credits, charged per second of output video:
| Model | 480p | 720p |
|---|---|---|
grok-imagine-video | 20 cr/s | 20 cr/s |
grok-imagine-video-1.5-preview | 32 cr/s | 32 cr/s |
480p and 720p cost the same. Choosing 480p saves you nothing.
Typical clips at the 6-second default:
| Clip | Classic | 1.5 |
|---|---|---|
| 6 s | 120 credits | 192 credits |
| 10 s | 200 credits | 320 credits |
| 15 s (text-to-video only) | 300 credits | 480 credits |
Credits are held at submission, before the job is dispatched. A request that the provider later rejects has already reserved its credits; failed and cancelled tasks are refunded automatically.
Underlying provider cost is $0.050/second for classic and $0.080/second for 1.5, both taken from xAI’s published model pages on 2026-07-21.
Limitations & gotchas
Section titled “Limitations & gotchas”- 1.5 hard-rejects reference images. Verified live 2026-07-17. There is no partial support and no fallback.
- 1.5’s aspect ratios are inherited, not verified. The 7 ratios offered on 1.5 were copied from classic. xAI publishes no ratio list for 1.5. Treat anything other than 16:9 on 1.5 as unverified — it may 400 at the provider.
- Over-spec resolutions silently downgrade. Asking for 1080p or 4K on either Grok model does not error — it returns 720p and bills at the 720p rate.
- The 15-second slider does not apply to every mode. Only text-to-video accepts up to 15 seconds. Image-to-video, reference-to-video and extension are capped at 10 seconds server-side and reject longer requests.
- Video edit is not reachable. Some Playcut surfaces still advertise video edit on classic; the API rejects it with “Video edit is only supported by wan-2.7”.
- Moderation is prompt-side and opaque. xAI exposes no strict-moderation toggle. A flagged job surfaces only as a failed task, with no category detail.
- REST API constraints are tighter than the picker in three places (verified in the tool request schemas on 2026-07-21). Every video tool shares one
aspectRatioenum (16:9,9:16,21:9,4:3,1:1,3:4) and one duration floor, so over the API:durationSecondsmust be 4 or above,3:2and2:3are not accepted on any video endpoint, and the extension endpoint locksdurationSecondsto 8. Those combinations are reachable only in the Playcut app, not through the public tools API.
How to generate
Section titled “How to generate”Both versions run through the standard tools registry endpoint described in the Quickstart. Select the version with the model field.
curl -X POST https://api.playcut.ai/api/v1/tools/generate-video-from-text/execute \ -H "Authorization: Bearer $PLAYCUT_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "workerId": "65f3a1b2c3d4e5f600000001", "model": "grok-imagine-video", "prompt": "A neon-lit alley in the rain, slow dolly forward", "aspectRatio": "16:9", "resolution": "720p", "durationSeconds": 6 }'curl -X POST https://api.playcut.ai/api/v1/tools/generate-video-from-image/execute \ -H "Authorization: Bearer $PLAYCUT_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "workerId": "65f3a1b2c3d4e5f600000001", "model": "grok-imagine-video-1.5-preview", "imageAssetId": "65f3a1b2c3d4e5f600000009", "prompt": "The subject turns toward camera and smiles", "resolution": "720p", "durationSeconds": 6 }'curl -X POST https://api.playcut.ai/api/v1/tools/generate-video-from-reference/execute \ -H "Authorization: Bearer $PLAYCUT_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "workerId": "65f3a1b2c3d4e5f600000001", "model": "grok-imagine-video", "referenceImageAssetIds": ["65f3a1b2c3d4e5f60000000a", "65f3a1b2c3d4e5f60000000b"], "prompt": "<IMAGE_1> walks through the set built in <IMAGE_2>", "aspectRatio": "9:16", "resolution": "720p", "durationSeconds": 8 }'curl -X POST https://api.playcut.ai/api/v1/tools/generate-video-extension/execute \ -H "Authorization: Bearer $PLAYCUT_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "workerId": "65f3a1b2c3d4e5f600000001", "model": "grok-imagine-video", "videoAssetId": "65f3a1b2c3d4e5f60000000c", "prompt": "The camera keeps pushing in as the lights cut out", "resolution": "720p", "durationSeconds": 8 }'Every call returns a taskId. Poll it with get-task-status or block on wait-for-task, then fetch the result with download-asset — the flow is identical for all models and is walked through in the Quickstart.
From an MCP client the same four operations are the tools generate-video-from-text, generate-video-from-image, generate-video-from-reference and generate-video-extension.