# More in this language: writing guide and additional exercises [← English](../../README.md) > [Illustrated prompts to copy and adapt](../../README.md#featured-prompts) — Browse the illustrated cases, or continue below for writing references and more prompts. ## Grok Imagine Video 1.5 capabilities These are official xAI API capabilities, not a list of SeaImagine interface controls. Editing and extension are companion workflows using a different documented model. See [model boundaries](../../docs/OFFICIAL.md). The table below is based on the current [xAI video generation documentation](https://docs.x.ai/developers/model-capabilities/video/generation) and [reference-to-video guide](https://docs.x.ai/developers/model-capabilities/video/reference-to-video). | Capability | Current behavior | Prompt-writing implication | | --- | --- | --- | | Text-to-video | Generates a first frame, then animates it | Describe the opening composition before describing motion | | Image-to-video | Uses one supplied image as the starting frame | Prompt mostly for movement, camera, atmosphere, physics, and sound | | Reference-to-video | Up to 7 reference images; up to 3 preset voices | Assign every input a role with `` and `` | | Duration | 1–15 seconds | Design one clear beat or a compact three-beat arc | | Resolution | 480p, 720p, and 1080p for text/image-to-video | Draft at lower resolution; finish important shots at 1080p | | Reference resolution | Up to 720p | Prioritize identity and product consistency over maximum resolution | | Aspect ratios | 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3 | Compose for the destination instead of cropping afterward | | Native audio | Audio track is generated by default | State ambience, Foley, dialogue, mix priority, and what must stay silent | | Companion workflow: video editing | Official examples use `grok-imagine-video`, not the 1.5 identifier | Check the separate model and provider support before using editing recipes | | Companion workflow: extension | See the separate [official workflow](../../docs/OFFICIAL.md) | Do not assume the SeaImagine 1.5 page exposes extension | > **Version note:** model behavior and API limits can change. This guide was reviewed on **September 24, 2026**. Verify production parameters against the official xAI documentation. ## The prompt formula Grok Imagine 1.5 usually responds better to a clear shot brief than to a bag of style keywords. Use this order: ```text MODE + OUTPUT State text-to-video, image-to-video, reference-to-video, edit, or extension. Add duration, aspect ratio, and intended delivery format. OPENING FRAME Describe subject, environment, composition, lens feeling, lighting, and palette. For image-to-video, say what in the supplied image must remain unchanged. ACTION + PHYSICS Give one primary action and two or three supporting motions. Describe weight, wind, water, cloth, particles, contact, or momentum only where relevant. CAMERA Use one motivated move: slow push-in, handheld follow, orbit, crane, pan, or locked frame. If the shot changes, give a simple time range for each beat. AUDIO Specify ambience, Foley, dialogue, music, and mix priority. CONTINUITY LOCK Repeat identity, wardrobe, product geometry, label, lighting, and spatial constraints. AVOID Name a short list of likely failures: extra limbs, geometry drift, unreadable text, camera teleporting, plastic skin, floating props, unwanted subtitles, or extra logos. ``` ### A compact reusable template ```text [Mode] Image-to-video, 8 seconds, 16:9. [Source lock] Preserve the supplied subject, composition, materials, colors, and lighting. [Action] The subject performs one clear action. Add subtle secondary motion in hair, fabric, reflections, particles, and background activity. Keep weight and contact believable. [Camera] One slow, continuous camera move with stable focus and no cuts. [Audio] Natural ambience and synchronized Foley; no music unless requested. [Continuity] No identity drift, object redesign, geometry warping, or new text. [Finish] End on a stable composition that can be used as the start of the next shot. ``` ## Multilingual prompts All fifteen homepages now contain eight illustrated cases with complete localized prompts, plus browser steps and community guidance. The table below lists additional inherited material. The English total remains 38 distinct recipes, including the eight homepage cases; translations do not add new scenarios. Chinese, Japanese and Spanish retain four extended recipes each, while the other eleven languages retain the workshop test scene in their writing guides. | Language | Localized guide | Additional material beyond the 8 homepage cases | | --- | --- | --- | | English | [English](../../README.md) | 30 category recipes | | 简体中文 | [简体中文](../../README.zh-CN.md) | [4 prompts](../../i18n/prompts.zh-CN.md) | | 繁體中文 | [繁體中文](../../README.zh-TW.md) | 1 shared test scene in the writing guide | | 日本語 | [日本語](../../README.ja-JP.md) | [4 prompts](../../i18n/prompts.ja-JP.md) | | 한국어 | [한국어](../../README.ko-KR.md) | 1 shared test scene in the writing guide | | Español | [Español](../../README.es-ES.md) | [4 prompts](../../i18n/prompts.es-ES.md) | | Français | [Français](../../README.fr-FR.md) | 1 shared test scene in the writing guide | | Deutsch | [Deutsch](../../README.de-DE.md) | 1 shared test scene in the writing guide | | Português (Brasil) | [Português](../../README.pt-BR.md) | 1 shared test scene in the writing guide | | Italiano | [Italiano](../../README.it-IT.md) | 1 shared test scene in the writing guide | | Русский | [Русский](../../README.ru-RU.md) | 1 shared test scene in the writing guide | | العربية | [العربية](../../README.ar.md) | 1 shared test scene in the writing guide | | Bahasa Indonesia | [Bahasa Indonesia](../../README.id-ID.md) | 1 shared test scene in the writing guide | | ไทย | [ไทย](../../README.th-TH.md) | 1 shared test scene in the writing guide | | Tiếng Việt | [Tiếng Việt](../../README.vi-VN.md) | 1 shared test scene in the writing guide | See the [language directory and localization rules](../../docs/LANGUAGES.md) for terminology, dialogue, locale, accessibility, and review guidance. For mixed-language projects, write shot direction in your strongest language, put spoken dialogue in quotation marks, and explicitly state language, regional variety, energy, pauses, and mix priority. ## API quick start This calls the official xAI API with a separate xAI account; it is not a SeaImagine API. Install `xai-sdk` and set `XAI_API_KEY` in your local environment. This example was checked against documentation, not executed as a paid generation. ```python import os import xai_sdk client = xai_sdk.Client(api_key=os.getenv("XAI_API_KEY")) response = client.video.generate( model="grok-imagine-video-1.5", prompt="A paper kite rises above sunlit rice terraces; slow crane up, natural wind and cloth physics, distant birds and soft string tension", duration=8, aspect_ratio="16:9", resolution="1080p", ) print(response.url) ``` For an image-to-video request, add `image_url`. For reference-to-video, use `reference_image_urls` and mention each asset in the prompt with numbered image tags. Generated URLs are temporary, so save production outputs promptly. See the official [video generation guide](https://docs.x.ai/developers/model-capabilities/video/generation) for the current SDK and REST examples. ## Quality checklist Before generating: - Pick one mode and one primary action. - Compose the first frame for the final aspect ratio. - Limit a 6–10 second shot to one camera move and one emotional beat. - Assign every reference image a single, explicit role. - Quote exact dialogue and keep it short enough for the duration. - State three to six continuity locks that truly matter. After generating: - Check faces, hands, labels, wheels, contact points, and reflections frame by frame. - Listen for dialogue timing, Foley synchronization, unwanted music, and abrupt audio tails. - Confirm the last frame is stable before extending the clip. - Change one variable per iteration; do not rewrite a successful prompt from scratch. ## Frequently asked questions ### What is the best Grok Imagine 1.5 prompt structure? Describe the opening frame, one clear action, one camera move, relevant physics, audio, continuity locks, and a short avoid list. For image-to-video, spend fewer words redescribing the image and more words on how it should move. ### Does Grok Imagine Video 1.5 support 1080p? Yes, current xAI documentation lists 1080p for text-to-video and image-to-video. Reference-to-video is currently capped at 720p. ### How long can a Grok Imagine 1.5 video be? The generation API currently accepts durations from 1 to 15 seconds. Longer narratives are best built as several shots with stable ending frames, then extended or edited together. ### How do I keep a character consistent? Use reference-to-video with clear, non-conflicting images: a neutral face/identity image, a full-body wardrobe image, and only the props or locations that matter. Assign each image a role and repeat the identity, wardrobe, and anatomy locks in the prompt. ### Should I include a negative prompt? There is no need for a giant keyword dump. Add a short `Avoid` sentence inside the main prompt that names the failures most likely for that shot: identity drift, extra fingers, object redesign, warped text, camera jumps, or new logos. ### Can Grok Imagine prompts be written in languages other than English? Yes. This repository includes localized guidance in 15 languages. For precise spoken dialogue, keep the quoted line in the target language and explicitly state the regional variety, delivery, pauses, and mix priority. Avoid silently translating dialogue inside an otherwise localized prompt. ## Contributing Original contributions are welcome. Read [CONTRIBUTING.md](../../CONTRIBUTING.md) before opening a pull request. Contributions may be original or properly attributed adaptations with permission. Include reproducible settings when tested, and only share media you have the right to use. ## Official resources - [Grok Imagine Video 1.5 announcement](https://x.ai/news/grok-imagine-video-1-5) - [Video generation documentation](https://docs.x.ai/developers/model-capabilities/video/generation) - [Image-to-video documentation](https://docs.x.ai/developers/model-capabilities/video/image-to-video) - [Reference-to-video documentation](https://docs.x.ai/developers/model-capabilities/video/reference-to-video) - [Video editing documentation](https://docs.x.ai/developers/model-capabilities/video/editing) - [Video extension documentation](https://docs.x.ai/developers/model-capabilities/video/extension) ## License Code and original written material in this repository are available under the [MIT License](../../LICENSE). Generated preview assets are provided for demonstration within this repository; review your own platform, model, and distribution terms before commercial use. --- If this library saves you time, star the repository and share a prompt improvement through a pull request.