How to Turn Text into Video with AI — Prompts That Actually Work
MyDreamVid · 2026-09-11
Text-to-video is exactly what it sounds like: you describe a scene in words, and a video model generates the footage. The gap between a muddy clip and a striking one is almost entirely in the prompt. Here is a structure that works, plus the practical details — length, cost, iteration — that tutorials usually skip.
A prompt structure that works
Cover four things, in roughly this order:
- Subject and action — who or what, doing what. "A paper boat drifts down a rain-soaked street."
- Camera — how the shot is filmed. "Low tracking shot following the boat, slow push-in at the end."
- Light and mood — the look. "Dusk, neon reflections on wet asphalt, cinematic, soft film grain."
- Sound (optional) — audio is generated too. Put dialogue in quotes and it will be spoken: a street vendor calls out "hot noodles, last pot tonight!"
You don't need magic words. You need concrete, filmable details — the model can't guess what "make it beautiful" means, but it knows exactly what "morning light sweeping across the table" looks like.
Choosing length, format and resolution
- Duration: 4–30 seconds on Seedance 2.5, 4–15 seconds on the 2.0 family. Shorter clips are cheaper and iterate faster.
- Aspect ratio: 16:9 for YouTube, 9:16 for Reels/TikTok/Shorts, 1:1, 4:3, 3:4, and 21:9 cinematic are all available.
- Resolution: 480p for cheap drafts, 720p as the sweet spot, 1080p for final renders. Note that 1080p output is H.265/HEVC — a few browsers can't play it inline, so download it and open it in a video player.
Iterate cheap, finish expensive
The single best cost habit: draft on Seedance 2.0 Mini or Fast until the prompt is right, then regenerate the final version on 2.5. A 5-second 720p draft on Mini is about 54 credits (≈ $0.54), so you can afford several attempts before committing to a high-end render. The exact cost of every generation is shown before you confirm — the number you see is the number you pay, and failed generations are refunded automatically.
If your prompt is declined
Two review layers can decline a prompt: the platform's content policy (no sexual content, minors, real people, political propaganda, hate, or violence) and the model's own review. In practice, harmless prompts that get declined usually pass after rewording — and Seedance 2.5 is stricter than 2.0 on identical wording, so a prompt refused on 2.5 sometimes passes on 2.0.
From one clip to a whole video
A single prompt gives you a single shot. For anything longer — an explainer, a story, a music-video sequence — write a shot list and generate it shot by shot; the film workspace automates exactly that, including joining the shots into one MP4.
Frequently Asked Questions
How long does a generation take?
Typically one to three minutes for a 5–10 second clip at 720p.
Is there a free trial?
There are no free credits, but there is also no subscription — top up from $10 (1,000 credits), and the credits never expire. Every job shows its exact price before you run it.
Can I use the videos commercially?
Yes. You own the videos you generate and can use them in commercial work, subject to the content policy and the rights you hold in anything you upload.
Can it generate speech in the video?
Yes. Dialogue written in quotes is spoken with natural intonation, even on the cheapest model. Audio adds nothing to the price on 2.0 Fast and Mini, and roughly doubles it on 2.5 and 2.0.