Canonical page: https://mixerai.org/en/knowledge/how-to/text-to-video/

[Mixer AI](https://mixerai.org/) › [Knowledge Base](https://mixerai.org/en/knowledge/) › [How to](https://mixerai.org/en/knowledge/how-to/) › How to create a video from text

# How to create a video from text

Without a starting photo, the prompt defines both appearance and movement. Begin with one clear episode. A short clip is easier to direct when its action has a beginning, development and visible ending.

[Try it on Mixer AI](https://mixerai.org/?dl=kb_text-to-video#/canvas)

## Step by step

Define a character, place and event: a courier approaches a door, puts down a box and rings the bell.

Add framing, direction and camera behavior. Separate multiple shots explicitly; do not ask for a locked camera and an orbit at the same time.

Select text mode, duration and sound, then check the price. Review action continuity, stable objects and whether dialogue fits the timing.

## Which model to use

[

Seedance 2.5

Build a short story, not just a moving picture: up to 30 seconds, recurring characters, your own references and sound.

](https://mixerai.org/en/knowledge/video/seedance-2-5/)[

Kling 3.0

Multiple shots, reference-based characters and optional sound for scenes that need more than an attractive camera move.

](https://mixerai.org/en/knowledge/video/kling-3-0/)[

Veo 3.1

A short scene with both picture and sound in mind, from rustling fabric to a character's line.

](https://mixerai.org/en/knowledge/video/veo-3-1/)

## FAQ

Must the prompt be in English?

Not every model requires English. Start with clear wording in your preferred language; compare a reformulation if it is misunderstood.

Why can a long prompt work worse?

It may contain competing actions and conflicting directions. Keep what can actually happen within the clip.

How do I reuse a character across videos?

Text may not preserve an exact face. Prepare a character image and use a reference-capable model.

[Try it on Mixer AI](https://mixerai.org/?dl=kb_text-to-video#/canvas)
