Studio lets you create AI-generated videos guided by your prompt, giving you more freedom and control over the creative process.
Here’s how to get started:
On the web, select Studio under Marketing tools in the left-hand menu. You can also open it from the Home page.
At the top of Studio, select Video. You can generate a video from reference frames, from a prompt, or from both.
Choose how you’d like to get started. There are three ways, and each one gives AI model a different level of freedom:
Click Start frame or End frame to open the image selector.
You can add:
Only a start frame to define how the video begins.
Only an end frame to guide its final shot.
Both frames to guide the transition between two compositions.
No frames if you want to build the video entirely from a prompt.
A start frame is useful when the subject, product, character, or opening composition must remain recognizable. An end frame helps when you already know how the final shot should look.
If you add both, choose images that can be connected through a believable movement or transition. Large changes in the subject, setting, camera angle, or visual style may produce an unstable result.
Use an image you created in Zeely if you wish. Choose one from:
Uploads: Images you previously uploaded.
Projects: Media saved in your Zeely projects.
Products: Images connected to your products.
Image ads: Images from the Image ads library.
Video ads: References available in Video ads.
Viral videos: References from the Viral videos library.
Avatars: Ready-made, custom, or liked avatars.
Or upload your own reference image. But keep in mind the requirements:
Maximum file size: 10 MB.
Supported formats: WEBP, PNG, JPEG, JPG, and HEIC.
When the selected model supports reference media, add the files through the media control near the prompt. Type @ in the prompt field to refer to a particular file.
For example:
Use @product_image for the product’s appearance. Use @reference_video for the slow camera movement and timing of the reveal. Keep the product packaging, colors, and logo unchanged.
Give each reference a clear purpose. Otherwise, the model may carry over a character, composition, setting, or mood that you did not intend to use.
You can also start from a prompt alone. For that, skip the starting and the end frames, and describe what you want to create.
Describe your video by typing your scene into the prompt box. You can describe a single scene, or connect several scenes by labeling them Scene 1, Scene 2, Scene 3, and describing what happens in each one.
Go into as much detail as you need, covering characters, actions, setting, camera movement, mood, or transitions between shots. A useful video prompt can cover:
The main subject
The action
Background movement
The setting
Camera position and movement
Lighting and visual style
Sound, when supported
Details that must remain unchanged
Instead of: Woman, city, rain, cinematic.
Write: A woman walks steadily toward the camera on a city street at night. She holds a black umbrella while her coat moves gently in the wind. Rain falls through the streetlights, trees sway behind her, and pedestrians cross the wet street in the background. The camera moves backward at the same pace, keeping her centered. Use cool blue lighting with warm reflections from shop windows.
Describe background movement as carefully as the main action. If the rain should fall, the trees should move, or people should cross the street behind the subject, include those instructions. Otherwise, parts of the scene may appear frozen or disconnected.
Plan the pacing. A short video cannot carry an unlimited number of cuts. Seven cuts in three seconds, for example, give the viewer less than half a second to understand each shot.
As a practical starting point:
Allow at least two seconds for each shot.
Use two to four shots for most ten-second videos.
Use no more than five shots in ten seconds, and only when every action is simple.
Give complex movements more time instead of breaking them into several rapid cuts.
Keep a continuous shot when a cut does not add anything useful.
These are more like creative guidelines than technical restrictions. Sometimes it might take four or five seconds to present a product, whereas a simple closing shot might take just two.
7. Use one clear action per shot. Give each shot one main action. Background movement and camera movement can support it, but they should not compete with it.
For example, a ten-second video showing a Roman soldier and a wooden barrier could be structured like this:
0 to 3 seconds: A Roman soldier approaches a closed wooden barrier. The camera follows him from the side while his cloak moves in the wind.
3 to 7 seconds: He stops, grips the wooden beam with both hands, and lifts it. Keep the camera steady and show the full movement.
7 to 10 seconds: He steps through the opening and looks back. The camera moves in slightly as the barrier remains raised behind him.
Each shot has one purpose: approach, lift, and pass through. This gives the model time to complete each movement and makes the sequence easier to follow.
Good to know: For longer prompts, add timecodes to show when each scene or action should happen. This helps guide the pacing and makes it easier for the model to follow the sequence you have in mind.
For example:
0 to 5 seconds: Establish the product on a marble counter in soft morning light. Steam moves gently in the background.
5 to 10 seconds: Move the camera closer as a hand reaches into the frame and picks up the product.
10 to 15 seconds: Show the person opening and using the product. Keep the label facing the camera.
15 to 20 seconds: Cut to a close-up of the product in the person’s hand. Light moves across the packaging.
20 to 25 seconds: Return the product to the counter and hold the final composition. Keep the logo clear and centered.
Make sure the timecodes match the duration selected in Studio. If you choose a ten-second duration, the final timecode should also end at ten seconds.
Timecodes guide the model, but they are not frame-accurate editing controls. Review the result and refine the prompt if an action begins too early, lasts too long, or does not finish clearly.
Also, describe cuts and transitions. This helps your video flow smoothly and feel like one cohesive piece.
For example:
Cut to a close up after the product opens.
Continue into the next shot without cutting.
Use a soft dissolve between the indoor and outdoor scenes.
Hold the final frame for two seconds.
Use cuts with restraint. One controlled camera movement will often look more convincing than several quick transitions competing for attention.
Sound is also important. It helps set the mood and bring your video to life. When the selected model supports audio and Sound is turned on, the model generates the sounds you describe in your prompt: background music, ambient sounds, or specific sound effects.
You can include:
Dialogue or narration
Environmental sound
Product sounds
Footsteps and physical movement
Music
Moments that should remain quiet
For example:
Add rain against the umbrella, quiet footsteps on wet pavement, and distant traffic. Keep the city ambience low. Do not add music or dialogue.
For a product video:
Add a soft click as the lid opens, followed by a light spray sound. Use quiet room ambience and subtle instrumental music underneath.
Avoid writing only “add sound effects.” Name the sounds and explain when they should occur.
If a product, person, logo, or other element must remain recognizable, say so directly.
For example:
Keep the bottle shape, label, logo, and colors unchanged. Do not add text to the background.
8. Open the AI model dropdown to compare your options:
Kling 4.0: Generates complete videos ranging from 3 to 30 seconds in a single generation, with native audio. It offers greater control over camera movement, scene composition, character movement, and transitions.
Kling 3.0: Delivers advanced realism with cinematic camera movement and native audio.
Seedance 2.0: Creates reference-guided motion with native audio.
Seedance 2.5: Offers faster generation with built-in audio support.
Seedance 2.5 Draft: Creates lower-cost 480p drafts that you can upgrade to HD within 7 days.
9. The available settings change depending on the model you choose. Each model card shows its default resolution, default duration, and credit rate per second for that resolution, so you can compare them before picking one.
Once you've chosen a model, fine-tune the settings underneath:
Aspect ratio: an aspect ratio may be selected by default based on the model, but you can choose another available option.
Resolution: available options depend on the AI model. For example, Seedance 2.5 Draft supports only 480p, Kling 3.0 only 1080p, and Seedance 2.0 offers 480p, 720p, 1080p, and 4K. You can only select resolutions that are available for the chosen model.
Duration: drag the slider anywhere to choose how long your video should be. Available durations depend on the AI model. For example, Kling 3.0 and Seedance 2.0 support videos up to 15 seconds, while Seedance 2.5, Seedance 2.5 Draft, and Kling 4.0 support videos up to 30 seconds. If your prompt uses timecodes, make sure the last one matches the duration you choose.
Sound: toggle on or off when supported by the selected model. When it's on, the model generates the sounds you describe in your prompt.
How are credit costs calculated?
Studio video models are priced per second of generated video. The cost depends on the AI model, the resolution, and the duration you select.
Each model card shows the rate per second for the resolution listed on it.
For example, a 10-second video with Kling 3.0 in 1080p costs 500 credits. A 10-second video with Seedance 2.5 costs 300 credits in 480p, 700 credits in 720p, and 1,500 credits in 1080p.
If you change the model, resolution, or duration, the amount displayed on the Generate button updates accordingly. While Zeely recalculates, a loading indicator may appear next to the previous amount. Wait for the updated cost before generating.
Important: In Studio, the prompt length doesn’t affect the cost.
The Generate button shows the total credit cost, and this is the amount deducted from your balance. When everything looks right, click Generate.
Save credits: test your idea with a moodboard first
Before you generate a video, you can preview it with a moodboard from the Zeely Creative Agent in Slack. It costs about 25 credits, much less than regenerating a finished video several times.
In Slack, mention @zeely in a channel or send Zeely a direct message. Describe the video you want: the idea, the mood, and the result you expect.
Attach your references to the same message. For example, your photo for a video with your avatar, a photo of the product for a product ad, or a photo of the exact model for a car video.
Ask Zeely to create a moodboard with 8–12 slides.
Review the slides and check that the scenes match your idea.
If something is off, reply in the same thread with what to change, and Zeely will update the moodboard.
Generate the video in Zeely only once you're happy with the moodboard. In the same thread, you can also ask Zeely for a ready prompt for the model you'll use.
Moodboard → review → adjust → approve → generate video. This way you control the result before you spend credits on the video itself.
Save credits: start with a draft
Not sure which idea will work best? Test it in 480p first, then pay for HD only for the videos you like.
In the AI model dropdown, select Seedance 2.5 Draft.
Set up your video as usual: add a start frame, an end frame, or reference images or videos, such as photos of your product, and describe the scene in the prompt. Then choose the aspect ratio, duration, and sound, and click Upscale. Draft videos are generated in 480p at a lower credit cost.
Once the draft is ready, you’ll see a Draft badge on the video card.
If you like the result, open the video and click Upscale in HD. You can also hover over the video card to find this button. The amount on the button shows how many credits the HD version will cost.
Zeely creates a new 1080p version of your video using Seedance 2.5. Your original draft stays in your library, so you’ll have both versions.
Good to know:
The HD version uses the same prompt, references, duration, aspect ratio, and sound settings as the draft. To change any of them, generate a new draft.
You can upgrade a draft to HD within 7 days of creating it. After that, the video stays available in 480p, but the Draft badge and the Generate in HD button disappear.
Each upgrade costs credits and creates a separate HD video.
Draft → review → upgrade to HD. This way, you can try several ideas for the price of one HD video.
10. Review, regenerate, or download your video. Once a video finishes processing, click on it to open the preview. From here, you can:
Play it back using the controls under the video and check the timestamp and total length.
Look over the details on the right, including the filename, when it was created, the full prompt you used (with any tagged references, like a start frame, called out inline), and the settings you generated it with (model, sound, duration, resolution).
Copy the prompt by clicking the copy icon next to "Prompt," which comes in handy if you want to reuse or tweak it for another generation.
Regenerate the video with the same prompt and settings if you want another take. This draws down credits again. The cost is based on the original video's model, resolution, and duration, and is shown on the button before you regenerate.
Download the file straight to your device.
Favorite or delete it using the star icon and the three-dot menu at the top of the panel.
A video can start with almost anything: a single image, an ending you already have in mind, or just an idea worth exploring. Try different prompts, reference frames, models, and settings, and see where they take you. See also How to create an image in Studio.
Have a question, or run into a problem? We're here. Reach our Support team at [email protected] or through the in-app chat.













