Every clip in this post was made from one sentence. No editing, no After Effects, no invite codes. We generated each one inside Writingmate, where Sora 2, Veo 3.1, Kling 3.0, Seedance 2.0, and PixVerse 5.5 all live in one $20/month plan. Here is the exact workflow with the actual prompts — the same format we use internally.
The 3-step workflow (same for every model)
- Describe your shot. Open the Video workspace and write one sentence. Details help: subject, action, lighting, camera style.
- Pick the model. Click the model chip and choose Sora 2, Veo 3.1, Kling 3.0, Seedance 2.0, or PixVerse 5.5. Set length (5–8s) and aspect ratio (Wide for YouTube, Portrait for Reels).
- Hit Create. Generation takes one to a few minutes depending on the model. Download the result or run a variation.
Sora 2 — cinematic motion, no invite
Best for: cinematic motion and realistic scene coherence. This is the model everyone wants to try — in Writingmate there is no invite code or waitlist.
The exact prompt used in the video below:
A hummingbird drinks from a red flower in ultra slow motion, water droplets sparkling in golden light, macro cinematic
Tip: Sora rewards physical detail — words like “ultra slow motion” and “macro” steer both the motion and the lens.
Veo 3.1 — native audio and sound effects
Best for: shots where sound matters. Veo generates native audio, dialogue, and sound effects together with the video.
The exact prompt used in the video below:
A street food chef tosses noodles in a flaming wok at a night market, sparks flying, cinematic slow motion close-up
Tip: mention sound sources in the prompt (sizzling, crowd noise) — Veo will score the scene to match.
Kling 3.0 — audio, voice control, multi-shot
Best for: multi-shot scenes and animating still images. “animate this” on a generated image is the classic Kling workflow for consistent characters.
The exact prompt used in the video below:
A surfer rides inside a massive glassy barrel wave at golden hour, spray glittering in the light, cinematic
Tip: Kling handles water and hair physics unusually well — great for sports and lifestyle shots.
Seedance 2.0 — real-world physics and camera control
Best for: cinematic realism with camera moves. It is also the model behind most of the viral animal edits you have seen lately.
The exact prompt used in the video below:
A lion cub is lifted toward the sunrise on a rocky outcrop as animals gather on the savanna below, epic golden light, cinematic
Tip: name the camera move (“slow push-in”, “orbit”) — Seedance actually follows it.
PixVerse 5.5 — fastest iterations, built-in effects
Best for: quick variation testing. When you want ten takes of an idea before lunch, PixVerse is the speed pick.
The exact prompt used in the video below:
An astronaut breakdances on the moon with Earth glowing behind, a disco ball floating in space, cinematic
Tip: PixVerse ships built-in effects — try the same prompt with different effect presets instead of rewriting it.
Which model should you pick?
| Model | Pick it when | Length |
|---|---|---|
| Sora 2 | You want the most cinematic single shot | up to 8s |
| Veo 3.1 | The scene needs sound or dialogue | up to 8s |
| Kling 3.0 | Animating images / multi-shot stories | 5–10s |
| Seedance 2.0 | Realistic physics and camera moves | 5–15s |
| PixVerse 5.5 | Fast variations and effects | 5–8s |
All five are included in the Pro plan (3 videos/month) and Ultimate (30/month), next to GPT-5, Claude, Gemini, and Grok for chat. One subscription instead of five.
Deep-dive guides per model
AI video model FAQ
Written by
Artem Vysotsky
Ex-Staff Engineer at Meta. Building the technical foundation to make AI accessible to everyone.
