繁體中文 · 한국어 · Français · Deutsch · Português · Italiano · العربية · Русский · Bahasa Indonesia · ไทย · Tiếng Việt · All 15 languages
A practical collection of 160 original Wan 3.0 AI video prompts for cinematic storytelling, product commercials, short-form drama, web series, social comedy, surreal UGC, VFX transitions, motion graphics, dialogue, localization, wildlife, industry, education, architecture, mobility, and production control.
| Browse 160 prompts | Submit a tested prompt | Add a translation | Open a pull request |
|---|
Practice continuous shots, product consistency and document/sheet/deck/webpage-to-video workflows. These new exercises are not the source videos' original prompts and have not been generation-tested. See the source guide for verification scope and limitations.
- Built around visible action and cause-and-effect, not keyword piles.
- Designed for text-to-video, image-to-video, start/end frames, reference-to-video, and video editing.
- Includes camera paths, timing, lighting, sound, identity locks, and relevant negative constraints.
- Works with multilingual visual descriptions and spoken dialogue.
- Organized with descriptive filenames and semantic headings for search and downstream gallery projects.
Useful outside submissions are actively welcome. If a Wan 3.0 prompt produced a result worth studying, use the guided prompt submission form. One submission should contain one complete prompt, its workflow and settings, a result link or screenshot where possible, and only assets you are allowed to share. Maintainers can then review, test, rewrite for consistency, and place it in the right production category.
You can also submit a pull request for a new scenario, a failure/success comparison, a more natural localization, or an accessibility improvement. Contributions do not need to be English-first: dialogue and on-screen language may use any language when the visual direction and speaker timing remain clear. Read the contribution guide before sending media or reference files.
Important
Wan 3.0 availability, duration, resolution, reference limits, and audio features may vary by region, product, or preview program. This repository does not invent fixed specifications. Use the controls shown by your current platform. Most templates can also be simplified for Wan 2.6 or 2.7.
[Output] duration + aspect ratio + visual medium
[Subject] reusable identity anchors + immutable details
[World] time + place + weather + spatial layers
[Action] trigger → continuous motion → visible result
[Camera] shot size + angle + one movement path + ending frame
[Look] light + palette + material + motion treatment
[Sound] ambience + action sound + music + dialogue, if supported
[Constraints] what must remain + the most likely failure modes
Minimal example:
8 seconds, 16:9, naturalistic cinematic footage. At dawn on a rain-washed old street, a young postal worker in a dark green coat rides a vintage bicycle through a shallow puddle. The front wheel parts the water into two low sprays; droplets fall back onto the stone instead of floating. The camera tracks from knee height at the rear-left, starts as a medium environmental shot, moves closer to the hand and bell, then rises toward warm light at the end of the street. Realistic speed and soft overcast light. Sound: tire through water, a distant shop shutter, one clear bicycle bell. Keep the face, coat, bicycle geometry, and travel direction consistent. No text, logos, extra limbs, jump cuts, or floating objects.
| Mode | Prioritize | Reliable instruction pattern |
|---|---|---|
| T2V | World, subject, action, camera | Describe one event before adding style |
| I2V | Motion and amplitude | Keep the input composition; specify exactly what moves and by how much |
| Start/end frames | Explainable transition | Describe the trigger, intermediate physics, and final state |
| R2V | Clear reference roles | “Image 1 = identity; Video 1 = motion rhythm only” |
| Video edit | One changed dimension | “Change only the weather; preserve action, timing, subject, and camera” |
| Audio/dialogue | Speaker timing | Use short lines; identify language, emotion, pause, and silent listener |
Read the detailed prompt engineering guide, model and platform compatibility notes, and troubleshooting handbook. These core guides are maintained in Simplified Chinese and include language-independent templates; the API cost guide below is in English.
Wan 3.0 is most useful when treated as a production family rather than a single text box. Start from text when the idea is still open, animate a designed keyframe when composition matters, edit an existing clip when only one dimension should change, and use reference-driven generation when identity or motion continuity is the priority. Across all four paths, the prompts in this repository emphasize temporal cause and effect, deliberate camera movement, stable subject anchors, physical motion cues, sound direction, and a short list of shot-specific constraints.
| Wan 3.0 path | Best starting material | Best use case | Prompting priority |
|---|---|---|---|
| Text to video | A written creative brief | Concept exploration, cinematic scenes, ad ideation | World → action → camera → look |
| Image to video | A strong first frame, plus an end frame where supported | Product shots, character shots, art direction | Preserve composition; define motion amplitude |
| Video edit | Existing footage | Weather, wardrobe, material, background, or mood changes | Change one dimension; lock everything else |
| Reference to video | Identity, style, audio, or motion references | Recurring characters and campaign consistency | Assign one explicit role to each reference |
For developers who want a simple hosted workflow, Flaq.ai exposes four available Wan 3.0 routes through a common asynchronous video-task API: Text to Video, Image to Video, Video Edit, and Reference to Video.
All four pages expose 2–30 second duration, 480p/720p/1080p output, five aspect ratios, optional sound, and seed-based iteration. Image to Video requires start and end frames; Video Edit accepts an existing clip up to 15 seconds. Reference to Video accepts up to 10 images, 5 video references, and 5 audio references; each video or audio reference can be 1–15 seconds. Its page also supports documents or public webpages as creative context.
| API | Current status | Recommended for |
|---|---|---|
| Wan 3.0 Text to Video | Available | Storyboards, ads, cinematic concepts, social clips |
| Wan 3.0 Image to Video | Available | Animating a first frame or directing a first-to-last-frame transition |
| Wan 3.0 Video Edit | Available | Natural-language changes to existing footage |
| Wan 3.0 Reference to Video | Available | Directing identity, style, motion, sound, and intent with images, clips, audio, documents, or webpages |
All four public pages currently show 10% OFF: $0.045/second at 480p, $0.09/second at 720p, and $0.18/second at 1080p, reduced from $0.05, $0.10, and $0.20 per second. That makes a practical cost ladder possible: prototype at 480p, validate motion and timing at 720p, then render only approved shots at 1080p.
Tip
10% OFF: The four Flaq.ai Wan 3.0 model pages currently display the discounted price, original price, and 10% off label directly. No coupon code is listed. Pricing and promotions can change, so recheck the selected model page before launching a large batch.
See the Flaq.ai Wan 3.0 API and cost guide for model selection, example budgets, integration notes, and a production checklist. Prices, availability, limits, and promotions can change; always verify the linked model page before production use.
Need one place to scan every title? Open the 160-scene master index.
| Category | Prompts | Includes | Open |
|---|---|---|---|
| X-inspired practical workflows | 10 | official video sources, continuous shots, identity, structured-brief workflows | Browse |
| Cinematic storytelling | 6 | drama, suspense, period scenes, one-take shots | Browse |
| Ads and products | 6 | beauty, food, technology, home, automotive | Browse |
| UGC, food and travel | 6 | vlogs, street food, stays, workshops, ASMR | Browse |
| Action and sports | 6 | pursuit, snowboarding, volleyball, parkour, VFX | Browse |
| Animation and fantasy | 6 | 2D, 3D, stop motion, East Asian fantasy, sci-fi | Browse |
| Music, comedy and social | 6 | live music, dance, rap, pets, office comedy, loops | Browse |
| Professional business and public service | 11 | SaaS, creator courses, podcasts, accessibility, telehealth, logistics | Browse |
| Education and science | 11 | climate, microscopy, safety, astronomy, marine science, museums | Browse |
| Architecture, hospitality and mobility | 11 | real estate, public space, accessible routes, e-bikes, transit, hotels | Browse |
| Production control and editing | 11 | green screen, previs, product rotation, local edits, reference roles, loops | Browse |
| Commerce, beauty, and retail | 10 | fit demos, shoppable video, skincare, packaging, accessible retail, catalog batches | Browse |
| People, dialogue, and localization | 10 | speaker turns, dubbing, sign language, podcasts, oral history, micro-drama | Browse |
| Nature, animals, and seasons | 10 | wildlife, animal care, weather, macro nature, seasonal change, observatories | Browse |
| Industrial and manufacturing | 10 | training, cobots, inspection, cold chain, digital twins, multi-SKU production | Browse |
| Short-form drama and web series | 10 | vertical drama, cliffhangers, prop continuity, episode variants, restrained dialogue | Browse |
| Social comedy and surreal UGC | 10 | phone video, immediate hooks, harmless surreal rules, punchlines, seamless loops | Browse |
| VFX, transitions, and motion graphics | 10 | matched geometry, controlled morphs, graphic-to-live action, local effects, rebuilds | Browse |
All 160 scenarios, characters, fictional products, and dialogue were created for this repository. No third-party preview image, watermark, promotional copy, or developer attribution is embedded in the collection.
Shoppable demonstrations, fit and texture comparisons, accessible product use, packaging continuity, consultations, and repeatable catalog campaigns. Open 10 prompts →
Clean speaking turns, multilingual dialogue, localized dubbing, sign-language framing, podcasts, documentary voiceover, and oral history. Open 10 prompts →
Patient wildlife observation, animal-care routines, macro physics, weather transitions, seasonal change, and non-invasive documentary direction. Open 10 prompts →
Safety rehearsals, cobot handoffs, facility explainers, inspection, cold-chain continuity, digital-twin overlays, and multi-SKU generation. Open 10 prompts →
Vertical cliffhangers, quiet mysteries, workplace comedy, public-service tension, serialized continuity, and repeatable episode endings. Open 10 prompts →
Phone-camera realism, instant hooks, one-rule surreal comedy, unboxing reversals, mockumentary timing, and seamless social loops. Open 10 prompts →
Blueprint-to-reality shots, material transitions, reference-aligned timelines, graphic-to-live-action explainers, local VFX edits, and exact rebuilds. Open 10 prompts →
Use one main language for the visual description and isolate exact dialogue:
Visual description: English cinematic production language.
Spoken dialogue: Mandarin Chinese, relaxed natural delivery.
Exact line: “今天的风，终于往海边吹了。”
No subtitles. The listener keeps their mouth closed and reacts with one small nod.
Avoid duplicating the entire prompt in multiple languages. Keep camera and material terminology in the main language; preserve only exact spoken or on-screen text in its target language.
The 15-language directory provides a localized prompt formula and the same complete comparison scene in every supported language: English, Simplified Chinese, Traditional Chinese, Japanese, Korean, Spanish, French, German, Brazilian Portuguese, Italian, Arabic, Russian, Indonesian, Thai, and Vietnamese.
- Build each short around one primary event.
- Reuse three to five identity anchors without paraphrasing them.
- Write motion as “first, then, finally.”
- Assign one main camera movement per shot.
- For I2V, describe motion more than static appearance.
- Show speed through water, dust, clothing, parallax, and sound.
- Lock product geometry, material, label, cap, and button positions.
- Keep dialogue short enough for natural pauses.
- Use three to six scene-specific negative constraints.
- Add one complex variable per iteration.
Original prompts, tested comparisons, and natural localizations are welcome. Use the prompt submission form for one tested recipe, or open a pull request for a larger category or translation. Do not submit copied prompt libraries, watermarked media, unauthorized celebrity likenesses, protected characters, or imitation of a living artist’s style. See CONTRIBUTING.md.
Repository code and original text are provided under the MIT License. Generated cover images are original project previews. This independent community resource is not an official model-provider publication.
Flaq.ai currently offers an affiliate program for developers, creators, educators, reviewers, and teams that recommend its AI model APIs and creative services. Sign in with a regular Flaq.ai account, complete the affiliate profile and agreement, generate your own referral link, and share that link in tutorials, reviews, communities, products, or websites.
- Earn 20% on a referred user's first valid paid order.
- Earn 10% on following valid paid orders made within the 60-day attribution window after registration.
- Refunds, chargebacks, risk-cancelled orders, attribution status, and policy abuse are reviewed before commission becomes payable.
Program rules may change. Review the current affiliate page and agreement before promoting Flaq.ai. The link above is the official program page, not this project's referral link.