Best AI Prompts for Text-to-Video Generation (2026)
What Runway and Google document about prompts, limits and costs
How to use: click Copy, replace the [amber fields] with your own material, and read the check under each set.
Prompts for text-to-video#
Five starting points#
5 promptsAct as a master cinematic director. Write a high-fidelity text-to-video prompt for a text-to-video model depicting a seamless drone shot soaring through a glowing fiber-optic neural network core with dynamic particle physics.Act as an expert commercial videographer. Create a precision text-to-video prompt for a sleek product reveal of a titanium smart device rotating slowly on a reflective obsidian surface under dramatic studio rim lighting.Act as a visual effects supervisor. Design a complex text-to-video generation prompt showing a futuristic metropolis transforming in real-time as autonomous drone swarms construct glass architectural structures.Act as an artistic motion designer. Draft a fluid, hypnotic text-to-video prompt featuring abstract liquid chrome waves merging and crystallizing into geometric typography against a dark minimalist background.Act as a documentary filmmaker. Write an atmospheric text-to-video prompt capturing misty morning light piercing through an advanced high-tech greenhouse managed by autonomous robotic arms.Limits to plan around (as documented on 2026-10-21)#
| Tool | Documented limits |
|---|---|
| Runway Gen-4.5 | Prompt up to 1,000 characters; 1280:720 or 720:1280; 2 to 10 seconds; 12 credits per second |
| Google Veo | 8-second videos at 720p, 1080p or 4K with native audio; durations of 4, 6 or 8 seconds; 16:9 or 9:16; Veo 3.1 Preview is listed for shutdown on October 22, 2026 |
| Google Gemini Omni Flash | Described as Google's default model for video generation; defaults to several shots unless you ask for a single continuous shot |
| OpenAI Sora 2 | Shut down on September 24, 2026; guide kept for historical reference |
For stills and image-to-video workflows, see our guide to image-to-image and video workflows. The full prompt library is on our prompt library page.
Key context and caveats#
Video tools publish limits more readily than prompt formulas. Runway's API reference describes a promptText field of up to 1,000 characters that should describe in detail what should appear, ratios of 1280:720 or 720:1280, and a duration from 2 to 10 seconds for Gen-4.5. Runway's pricing page lists Gen-4.5 at 12 credits per second, with credits sold at $0.01 each, which works out to $0.12 per second by our arithmetic. We found no Runway prompt-writing guide for Gen-4.5, so we do not claim one.
Google publishes more prompt advice. Its Veo guide says good prompts are descriptive and clear and lists subject, action, style, camera positioning and motion, composition, focus and lens effects, and ambiance. Its Omni guide notes the model tries to create a few different shots by default. OpenAI's Sora guide is retained for historical reference only.
At a Glance#
| Question | Answer from provider pages |
|---|---|
| What do prompts need? | Subject, action, style, camera, composition (Google) |
| How long can a Runway prompt be? | 1,000 characters |
| How long can a Runway clip be? | 2 to 10 seconds |
| Is Sora 2 available? | No; shut down on September 24, 2026 per OpenAI |
Sources & Further Reading#
Every source below was fetched on 2026-10-21 unless a row says otherwise.
| Claim | Source | Section or date |
|---|---|---|
| promptText up to 1,000 characters; 1280:720 or 720:1280; 2 to 10 seconds | Runway API reference | Text to video, model gen4.5; undated |
| Gen-4.5 at 12 credits per second; $0.01 per credit | Pricing (Runway API) | Video Generation Pricing; undated |
| Veo prompt elements; 8-second videos; durations; aspect ratios | Veo (Google Gemini API) | Prompt writing basics; API parameters; undated |
| Veo 3.1 Preview shutdown on October 22, 2026; Omni Flash as default | Video generation (Google Gemini API) | Intro and Veo 3.1 notice; undated |
| Omni prompt guidance and single continuous shot | Gemini Omni (Google Gemini API) | Text to video; prompt guide; undated |
| Sora 2 and Videos API shut down on September 24, 2026 | Video generation with Sora (OpenAI) | Page opening notice; undated |
Direct answer
What are the best AI prompts for text-to-video generation?
Prompt guidance for video comes from each provider, and availability changes fast. Google advises describing subject, action, style, camera and composition, and Runway limits Gen-4.5 prompts to 1,000 characters and clips to 2 to 10 seconds. Check current limits and costs before you plan a project.
How we sourced this
Model names, features and limits change often. The prompts below are our own untested starting points, and we did not measure which wording performs best. Facts are as printed on each provider page on the date shown. Video model availability changes quickly: Google lists a Veo 3.1 Preview shutdown on October 22, 2026, and OpenAI says Sora 2 shut down on September 24, 2026.
Need a different task?
Browse all 36 tasks and 180 prompts, with search and level filters.
How we weigh evidence, and what terms like AGI and ASI mean on this site, is explained in our methodology.
Frequently asked questions
What are the best AI prompts for text-to-video generation?
Prompt guidance for video comes from each provider, and availability changes fast. Google advises describing subject, action, style, camera and composition, and Runway limits Gen-4.5 prompts to 1,000 characters and clips to 2 to 10 seconds. Check current limits and costs before you plan a project.
What should a text-to-video prompt include?
Google lists subject, action and style as the basics, with camera positioning and motion, composition, focus and lens effects, and ambiance as optional elements. For its Omni model it advises scene description, camera movement, lighting and mood, and a phrase such as in a single continuous shot to avoid cuts.
Is Sora still available for AI video?
OpenAI says the Sora 2 models and Videos API were shut down on September 24, 2026 and that no one-to-one replacement API is available. Runway Gen-4.5 and Google video models were documented as available on the date we checked, but confirm current status before building on them.