Vertical video AI for TikTok, Reels and Shorts
To make a TikTok, Reels or YouTube Shorts video with AI, open the Video section, set the 9:16 aspect ratio and describe the scene in text, upload a photo as the first frame or attach references. Kling, Seedance, Veo, WAN and MiniMax H3 all render vertical frames. Kling 3.0, Kling 3.0 Omni and WAN 3.0 make sound in the same run, and you can add a voiceover, Suno music or lip sync separately.
Updated:
Models for vertical 9:16 clips
| Version | Developer | Modes | Duration | Resolution | Aspect ratios | Sound | References | Prompt | Features | Price from |
|---|---|---|---|---|---|---|---|---|---|---|
| Kling 3.0 | Kling | text to video, image to video, first and last frame | 3-15 s | 720p, 1080p, 4K | 16:9, 9:16, 1:1 | on/off switch | no references | up to 2,500 characters | from 20 tokens | |
| Kling 3.0 Omni | Kling | text to video, image to video, first and last frame, reference images, video edit | 3-15 s | 720p, 1080p, 4K | 16:9, 9:16, 1:1 | on/off switch | up to 7 images, up to 1 videos | up to 2,500 characters | from 20 tokens | |
| Seedance 2 | ByteDance | text to video, image to video, first and last frame, reference images, video edit | 4-15 s | 480p, 720p, 1080p, 4K | 9:16, 16:9, 1:1, 3:4, 4:3, 21:9 | on by default | up to 9 images, up to 3 videos | accepts an audio reference; video references work in text to video mode; 21:9 available | from 23 tokens | |
| Seedance 2 Fast | ByteDance | text to video, image to video, first and last frame, reference images, video edit | 4-15 s | 480p, 720p | 9:16, 16:9, 1:1, 3:4, 4:3, 21:9 | on by default | up to 9 images, up to 3 videos | accepts an audio reference; video references work in text to video mode; 21:9 available | from 18 tokens | |
| Seedance 2 Mini | ByteDance | text to video, image to video, first and last frame, reference images, video edit | 4-15 s | 480p, 720p | 9:16, 16:9, 1:1, 3:4, 4:3, 21:9 | on by default | up to 9 images, up to 3 videos | accepts an audio reference; video references work in text to video mode; 21:9 available | from 12 tokens | |
| Seedance 2.5 | ByteDance | text to video, image to video, first and last frame, reference images, video edit | 4-30 s | 480p, 720p, 1080p | 9:16, 16:9, 1:1, 3:4, 4:3, 21:9 | on by default | up to 30 images, up to 10 videos | accepts an audio reference; video references work in text to video mode; 21:9 available | from 33 tokens | |
| Veo 3.1 Fast | text to video, image to video, first and last frame | 8 s | 720p, 1080p, 4K | 9:16, 16:9, 1:1, 3:4, 4:3 | no switch | up to 1 images | up to 1,000 characters | from 63 tokens | ||
| Veo 3.1 Lite | text to video, image to video | 8 s | 720p, 1080p | 9:16, 16:9, 1:1, 3:4, 4:3 | no switch | up to 1 images | up to 1,000 characters | from 32 tokens | ||
| Veo 3.1 Quality | text to video, image to video, first and last frame | 8 s | 720p, 1080p, 4K | 9:16, 16:9, 1:1, 3:4, 4:3 | no switch | up to 1 images | up to 1,000 characters | from 250 tokens | ||
| WAN 3.0 | Alibaba | text to video, image to video, reference images | 2-30 s | 480p, 720p, 1080p | 9:16, 16:9, 1:1, 3:4, 4:3 | on by default | up to 10 images, up to 5 videos | up to 20,000 characters | accepts an audio reference | from 8 tokens |
| WAN 3.0 Prime | Alibaba | text to video, image to video, reference images | 2-30 s | 480p, 720p, 1080p | 9:16, 16:9, 1:1, 3:4, 4:3 | on by default | up to 10 images, up to 5 videos | up to 20,000 characters | accepts an audio reference | from 11 tokens |
| MiniMax H3 | MiniMax | text to video, image to video, reference images | 5-15 s | 768p, 2K | 9:16, 16:9, 1:1, 3:4, 4:3 | no switch | up to 9 images, up to 3 videos | picks the mode from your attachments | from 24 tokens |
How the models differ
On Kling 3.0 and Kling 3.0 Omni, sound from the generation has a switch, and on WAN 3.0 it is on by default. Seedance, Veo and MiniMax H3 have no sound switch, but Seedance and WAN 3.0 accept a ready track as an audio reference.
Veo 3.1 has a fixed length, the other models let you pick one within a range, and Seedance 2.5 and WAN 3.0 make the longest clips per run. The resolution ceiling varies by version, from 720p to 4K, and the exact numbers are in the table.
Kling 3.0 works with text and frames but takes no references. Kling 3.0 Omni, Seedance, WAN 3.0 and MiniMax H3 accept several images and videos, while Veo 3.1 takes one image. Kling Motion Control is not in the table because it takes the aspect ratio from the source clip: if the motion is already filmed vertically, it moves your character the same way.
How to make a TikTok or Reels clip with AI
Set 9:16
In the Video composer, pick the vertical aspect ratio before you run. The frame is built for a phone screen from the start, so there is no horizontal clip to crop later.
Choose what to start from
If the scene does not exist yet, text is enough. When the product or person has to look a certain way, upload a photo as the first frame, and for a recognizable character or style, attach references.
Describe the first second
Open the prompt with what the viewer sees right away, such as a close-up of the product in a hand or a sharp camera turn. At the end, say how the clip finishes.
Set the length and run
Pick a duration within the model range and turn on sound where there is a switch. The token price is shown by the run button, and the finished clip downloads as a file.
What works in a short video
People scroll fast, so the main thing has to be in the frame from the very start: the product, a face, the result or an unexpected action. Save long intros and logo openers for other formats.
One clear action per clip is usually enough if the camera is alive, say a push-in or a walk past the subject. A close-up reads better than a wide shot on a small screen, and in a narrow vertical frame it is easier to stack objects top to bottom than to line them up.
On-screen text repeats the main point for people watching with the sound off. It is easier to add when you post, in the built-in editor of the app, where you can fix or translate it. Leave room for it, such as a calm background at the top of the frame, and keep key details near the center, because the app draws its own buttons and captions along the edges.
Plan the sound before you generate. If the voice or music is ready, match the clip length to the track, not the other way around.
A series of clips for one product
A series needs one reference image of the product: a real photo or a 9:16 frame from the Photo section. The To video button on an image sends it to the composer as the first frame. From clip to clip, change the scene and the opening, and leave the product itself alone.
If the same person appears in every clip, create them in the Characters section and pick them in the composer, so the first frame is built with their face. To try several openings at once, use the Queue button: it sends a few different prompts in one go, with the model, aspect ratio, duration and references shared.
Sound, voice and music
- Sound from the generation: a switch on Kling 3.0 and Kling 3.0 Omni, on by default on WAN 3.0.
- Voiceover: ElevenLabs, Minimax Speech-02 HD and Seed Audio 1.0 read your text in the Voiceover section.
- A person who speaks on screen: a finished voiceover has a button that sends it to Lip sync. Add a clip or a photo there, and the lips follow the speech.
- Music: Suno in the Music section writes a track from a description, with or without vocals. The To video button moves it to the composer, where you can attach it to Seedance or WAN 3.0 as an audio reference.
Getting the clip ready to post
Aspect ratio and duration are set at generation time, so before you run, decide where the clip will go and how long it needs to be, then pick a model that covers that length. Platform requirements change over time, so check them in the TikTok, Instagram and YouTube help pages.
If the clip needs more detail, the Post-processing section upscales it to 1080p, 2K or 4K. Before posting, watch the clip on a phone to see whether anything important hides under the buttons. Also check how the platform asks you to label videos made with AI.
More tools for short videos
Where to find it in ClickBait
Videos for TikTok and Reels lives in ClickBait under the Video section. Sign up, pick a version from the model list and describe what you want to get.
The token price is shown on the run button before you start. If a generation fails, the tokens go back to your balance automatically.
New accounts created with Google or email get a welcome bonus of 10 tokens. These are promo tokens and expire after 14 days.
Models for this task
Kling
video 3-15 s; matches the source video, up to 4K
images up to 4K
up to 10 references
from 2 tokens
Seedance
video 4-30 s, up to 4K
from 12 tokens
Veo
video 8 s, up to 4K
from 32 tokens
WAN
video 2-30 s, up to 1080p
up to 4 references
from 2 tokens
MiniMax H3
video 5-15 s, up to 2K
from 24 tokens
Other tools
AI video generator
12 models
Kling, Seedance, Veo, WAN, MiniMax H3
from 8 tokens
Image to video
13 models
Kling, Kling Motion Control, Seedance, Veo, WAN, MiniMax H3
from 8 tokens
AI image generator
18 models
Kling, Nano Banana, GPT Image, Seedream, FLUX, WAN, Z-Image
from 2 tokens
AI voiceover
5 models
ElevenLabs, Minimax Speech-02 HD, Seed Audio 1.0
from 2 tokens
Lip sync
5 models
Lipsync 2, Lipsync 2 Pro, Lipsync 3, InfiniteTalk, LatentSync
from 4 tokens
AI music generator
9 models
Suno
from 5 tokens
Video upscaler
1 model
Upscale
from 2 tokens
AI UGC video
12 models
Kling, Seedance, Veo, WAN, MiniMax H3
from 8 tokens
Product ad video
12 models
Kling, Seedance, Veo, WAN, MiniMax H3
from 8 tokens
Creatives for affiliate marketing
30 models
Kling, Seedance, Veo, Nano Banana, GPT Image, Seedream, FLUX, WAN, MiniMax H3, Z-Image
from 2 tokens
Marketplace product photos
17 models
Kling, Nano Banana, GPT Image, Seedream, FLUX, WAN
from 2 tokens
Ad posters and banners
17 models
Kling, Nano Banana, GPT Image, Seedream, FLUX, WAN, Z-Image
from 2 tokens
Frequently asked questions
Which AI makes TikTok videos?
Every model in the table supports 9:16: Kling, Seedance, Veo, WAN and MiniMax H3. For sound straight from the generation, use Kling 3.0 or WAN 3.0; for a longer clip, Seedance 2.5 or WAN 3.0; for a recognizable character, a model with references: Kling 3.0 Omni, Seedance, WAN 3.0 or MiniMax H3.
Can I make YouTube Shorts from text?
Yes. Pick 9:16, describe the scene and run: every model in the table works from text. Use the first sentence of the prompt for what the viewer should see in the first second.
Can the AI make a Reels clip from a photo?
Yes, upload the photo as the first frame and describe the motion. It is better to start from a vertical photo, since it matches the clip format. You can make a suitable frame in the Photo section and send it over with the To video button.
How do I add voice and music to a clip?
Make the voice in Voiceover and the music in Music with Suno. The To video button sends a track to the video composer as an audio reference for Seedance and WAN 3.0, and a voiceover goes to Lip sync, where the lips of the person on screen follow the speech.
How do I make a series of clips for one product?
Use the same product photo as the first frame or a reference in every clip, and change the scene and the opening. The Queue button runs several versions at once: the prompts differ, while the model, aspect ratio and duration are shared.
How long will the clip be?
You pick the length before you run, within the limits of the model. Veo 3.1 has a fixed 8 seconds, Seedance 2.5 and WAN 3.0 go up to 30, and the rest top out at 15 seconds. The range for each version is in the table.