Generating Videos

Generating videos

Mavic can generate short-form videos for Reels, Shorts, TikTok, ads, and product demos — from a text prompt, from a single image, or from two images that act as the first and last frame.

This article covers how each method works, which models support which mode, and how to keep videos on-brand using @-mentions.

Three ways to generate a video

Method	What it does	When to use
Text-to-video	You describe the video in words and Mavic generates it	You're starting from scratch — no source image
Image-to-video (animate)	You give Mavic one image and it brings it to life	You already have a strong still and want gentle motion
Image-to-video (first & last frame)	You give two images — Mavic morphs between them	You want a directed transformation (open → close, day → night, before → after)

Heads up: the first-and-last-frame mode is only supported by select models. See Which model does what below.

How to start

You have two entry points.

1. From a chat prompt

Open a chat and type what you want.

Text-to-video:

Generate a 6-second vertical video of a coffee cup being poured against a sunlit kitchen window, soft natural light, in @Brand Colors.

Image-to-video (animate):

Animate @Image: spring_collection_hero — gentle camera push-in, soft fabric movement, 5 seconds, vertical 9:16.

Image-to-video (first & last frame):

Generate a video that morphs from @Image: closed_box to @Image: open_box_with_product. Smooth reveal motion, 4 seconds, square.

2. From the Video Generation workflow

Click the workflow icon in the chat composer and pick a video workflow. The form asks for:

Mode — text-to-video / animate image / first-and-last frame
Subject and action — what's in the video, what's happening
Style — photorealistic, cinematic, anime, illustrated, product-shot
Camera motion — static, slow zoom-in, push-in, pan, tracking
Mood / lighting — golden hour, studio, moody, bright
Duration — typically 4–8 seconds for short-form
Aspect ratio — 9:16 (Reels/Shorts/TikTok), 1:1, 16:9
Source image(s) — required for image-to-video modes
Model — leave on Auto, or pick a specific one

Mavic builds a polished prompt from your inputs and generates the video.

Using `@` to keep videos on-brand

The @ system is what makes Mavic videos feel yours and not stock-AI. Tag in any of these:

@Image — the still you want animated, or your first/last-frame inputs
@Logo — to overlay your brand mark
@Product — so Mavic grounds the video in real product specs (color, material, packaging)
@Customer profile — Mavic adapts setting and casting to match your audience
@Writing style — applied to any on-screen text, captions or voice-over scripts
@Template — for branded intros / outros if you've saved one
@Social post — to extend an existing post into video form

Strong prompt example:

Animate @Image: founder_portrait with a slow push-in camera. Add the @Logo bottom-right at 60% opacity. Voice-over script in @Writing Style: Founder Voice. 9:16, 8 seconds.

Which model does what

Mavic includes several video models. Auto picks the best one for your prompt — leave it on Auto unless you have a reason to override.

Model	Strengths	Modes supported
Google Veo 3.1-Fast	Fast generation, great for short-form social	Text-to-video
Seedance 1.0 Lite / 1.5 Pro (ByteDance)	High visual quality, strong realism	Text-to-video, image-to-video (animate)
Kling v2.6 / v3 Omni	Best for directed motion and complex transformations	Text-to-video, image-to-video (animate), first-and-last frame

If you specifically need first-and-last frame morphing, pick Kling v3 Omni in the model dropdown.

Tips for stronger video output

Describe motion explicitly. "Slow push-in" beats "dynamic camera". Generic terms produce generic motion.
Keep one clear action per video. Trying to fit three things into 6 seconds reads as chaos.
Pair with a script. Use a chat prompt to generate the voice-over or on-screen text, then @-tag your Writing style.
Plan for sound-off. 85% of social video is watched muted — add captions or on-screen text in the post editor.
Generate multiple takes. Ask for 3 variations and pick the best — generation is fast.
Use square or vertical for social. 9:16 for Reels/Shorts/TikTok, 1:1 for feed.

Aspect ratios

Aspect	Best for
9:16 (vertical)	Instagram Reels, TikTok, YouTube Shorts, Stories
1:1 (square)	Instagram feed, LinkedIn
16:9 (landscape)	YouTube, in-stream ads, web embeds

Where generated videos go

Every video lands in Library → All Assets with version history. From there you can:

Download — MP4
Use in a post — attach to a Reel, Story, or Short draft
@-tag in a future prompt to extend, restyle, or repurpose
Restore previous versions — Mavic keeps a history of every refinement

Refining a generated video

After Mavic produces a video, you can:

Type follow-up changes in the Add feedback box: "Make the camera slower and brighten the highlights."
Re-roll the same prompt for new takes
Compare versions side-by-side
Restore an earlier version

See Editing & versioning generated assets.

Common questions

My first-and-last-frame video looks distorted.

The two source images need to share roughly the same composition. If frame 1 is a close-up and frame 2 is a wide shot, the morph will warp. Either re-shoot with similar framing or describe the transition explicitly: "Slow zoom-out from close-up to wide."

Can I generate videos longer than 8 seconds?

Most models are tuned for 4–8 second clips. For longer videos, generate multiple clips and stitch them together in your editor. Mavic can also generate a continuation that starts from the last frame of a previous clip.

Why is my video taking so long?

Video generation is more credit- and compute-heavy than images. Expect 30 seconds to a few minutes per clip depending on model and length.

Updated: May 2026

Updated on: 07/05/2026

Was this article helpful?

Thank you!