Creatify Boreal: AI video with sound, cut to the second

Creatify Boreal has arrived in Picsart AI Playground, and it writes video and sound together from one description. Ask for a clip of one second or twenty, or any whole second between, at resolutions reaching 2K.

That duration range is the part worth pausing on. Most video models hand you a short list of fixed lengths and leave you trimming afterwards. Boreal lets you ask for six seconds because the cut needs six seconds, or one second because you are building a hook.

Here is what the model does, the controls behind it, how its prompts are structured, and the kind of work the exact-length approach suits.

What Creatify Boreal is

Boreal is a text-to-video model built for product, UGC, and presenter clips. You describe a scene and it returns finished video with matching sound in a single pass, so ambience, motion, and audio arrive together rather than being assembled afterwards.

It takes three kinds of input:

• A text prompt. The required one, up to 5,000 characters, which is room for a full shot description rather than a line.
• A reference image. One image to anchor what a subject, product, or setting looks like.
• An audio track. Your own sound, supplied instead of leaving the model to generate it.

A negative prompt sits alongside them for the things you want kept out of the frame.

Every control it gives you

The floor is one second, and every step above it is available. No other video model in the Picsart catalog will generate something that brief. A one-second generation is a real output, not a preview setting.

Auto is an aspect ratio option. Leave it alone and the model picks a framing that fits what you described, which is useful before a placement is decided.

720p is the default and 2K is the ceiling. The gap between them is wide enough that drafting at the default and finishing at 2K is a sensible habit.

The disclosure flag is yours to set. Turn it on and the output carries an AI-generated marker, which matters for platforms and clients that ask for one.

Write the prompt in blocks

The example prompt Picsart publishes for Boreal is not a paragraph. It is four labelled blocks, one for each layer of the clip.

A product ad in four blocks

[VISUAL] A cinematic product close-up of a sparkling-water can on wet black stone. Condensation rolls down the can as the camera makes a slow half-orbit. [SPEECH] “Clean flavor. Bright finish. Nothing extra.” [SOUNDS] A crisp can opening, gentle fizz, and restrained electronic music. [TEXT] None.

Copied

`[VISUAL]` carries the shot and the camera move. `[SPEECH]` holds the line to be spoken. `[SOUNDS]` covers ambience and music. `[TEXT]` is set to None when nothing should appear on screen.

Picture and sound come out of that single prompt, so the block describing the audio is doing as much work as the one describing the frame. Written this way, a 5,000-character allowance stops looking generous and starts looking necessary.

Why exact lengths matter more than they sound

Short-form video is cut to fixed slots. A pre-roll is six seconds. A hook is two or three. A product beat inside a longer edit is however long the shot beside it leaves room for.

When a model only offers five or ten, every one of those jobs ends in the timeline, trimming a clip down and losing whatever the model put at the end of it. Motion that was paced across ten seconds does not read the same cut to six.

Generating at the target length instead means the pacing you get is the pacing you keep. The model composes the whole move inside the duration you asked for, so the beat lands where the edit needs it rather than where the trim happens to fall.

A way of working that fits it

1. Draft at 720p and short. Two or three seconds is enough to see whether the subject, framing, and look are right. Nothing later depends on resolution.
2. Settle the aspect ratio early. It shapes the composition, so changing it later means regenerating rather than cropping. Use auto only while the placement is still open.
3. Add the reference image once the scene works. Bring in a product or a face when the shot around it is already behaving, so you can tell which change did what.
4. Generate at the length the edit needs. Not a round number, the real one. This is what the per-second range is for.
5. Finish at 1080p or 2K. Raise resolution last, and only on the take you are publishing.

The sequence puts every cheap decision ahead of every expensive one.

What to make with it

Boreal’s three stated strengths point at specific formats:

• Product clips. A single item, one reference image, and a short move around it. Sound arrives with the motion instead of being tracked down later.
• UGC-style ads. The format that needs volume more than polish, where generating a dozen variations of one idea is the whole method.
• Presenter clips. Talking segments with synchronized audio, or your own recorded track supplied through the audio slot.
• Social hooks. One to three seconds, sized to stop a scroll, at a length most models cannot generate at all.
• B-roll to fill a gap. A specific number of seconds between two shots you already have.

Where Creatify Boreal sits in Picsart

The model appears by name in the video selector inside Picsart AI Playground. Choosing it exposes the controls above beside the prompt field, and nothing else has to be installed or configured first.

Boreal shares its canvas with image and music generation, which is what keeps a finished clip in the same project as the artwork and the track that accompany it. Jobs it was not designed for belong elsewhere in the Picsart AI models roster, now past 180 entries. Once the footage exists, editing and assembly happen in the Picsart AI video generator.

Start generating with Creatify Boreal

Head into Picsart AI Playground, choose Creatify Boreal, and run three seconds at 720p with the audio written into your description. Then run it again at whatever length your edit is really asking for.