Pick from Seedance, Veo, Kling, Wan, Grok, Hailuo and HappyHorse, then set start and end frames, 9:16 to 21:9 framing and up to 4K. Built for creators, marketers and filmmakers planning a flow AI video.
Three flow AI video input modes plus an image studio for first frames
Type the scene and match it to a model: Seedance 2.0 offers 480p up to 4K at 5, 10 or 15 seconds, and Kling V3 Turbo lists 720p or 1080p.
Add an opening still and a closing still, and Veo's Image to Video mode fills the motion between them. Grok Standard also takes an image, with 6, 10 or 15 seconds at 480p or 720p.
Need a starting still? The image studio lists 15 models, including Nano Banana 2 at 1K, 2K or 4K and Flux 2 Pro, so you can plan the first frame before the motion.
Guide Veo Fast or Veo Lite with a reference image when one subject should carry through a flow AI video series. The mode is 16:9 only, so plan widescreen framing.
You know the shot you want; the hard part is choosing a model that can frame it. A flow AI video starts with a written prompt. You can add a start and end frame, or a reference image on Veo Fast and Veo Lite (16:9 only). Each model lists its own options, such as ratio, length and resolution. In one flow AI video, a wide 21:9 shot and a 9:16 vertical can each use a different model.
Pick the flow AI video input that matches what you already have: words, two frames or a reference image.
Good when your flow AI video is still an idea with no images. Seedance 2.0 accepts prompts up to 20,000 characters with 21:9 and adaptive ratios, and Kling 3.0 offers std, pro and 4K modes.
Long Prompt Room
Seedance models take 20,000 characters, enough for a full shot list.
Wide Ratio Range
Pick 1:1, 4:3, 3:4, 16:9, 9:16 or 21:9 on Seedance.
Set Clip Length
Seedance 1.5 Pro offers 4, 8 or 12 seconds; Kling 3.0 offers 5, 10, 15.
Each flow AI video setting comes straight from the model list
Seedance, Veo, Kling, Wan, Grok, Hailuo and HappyHorse share one flow AI video picker, with each version listed by name, like Kling V3 Turbo.
Seedance 2.0 lists 480p, 720p, 1080p and 4K, Kling 3.0 has a 4K mode, and Hailuo 2.3 offers 768P or 1080P.
Seedance 1.5 Pro starts at 4 seconds, Grok Standard offers 6, 10 or 15, and Kling 2.6 keeps to 5 or 10.
Frame vertical, square or ultra-wide shots; Grok Standard adds 2:3 and 3:2, and Seedance 2.0 adds an adaptive ratio.
Prompt limits vary by model: 20,000 on Seedance, 5,000 on Wan, Grok Standard and HappyHorse, 2,500 on Kling 3.0, 1,000 on Kling 2.6.
Fourteen models list an Image to Video mode, from Wan 2.7 and HappyHorse 1.1 to Hailuo 02 Pro, so a photo can lead the shot.
Model first, words second, shot settings last
Veo Fast is preselected; switch to any of the other 17 if your flow AI video needs 4K, 21:9 or a photo input.
Describe subject, action, camera move and light, staying inside the model's limit, which ranges from 1,000 to 20,000 characters.
Choose ratio, duration and resolution for the shot, then press Generate and download the finished clip.
0 / 2000
Borrow a look as the starting point for a flow AI video prompt: product close-ups, wide landscapes, portraits and motion.
Real workflows: the input they bring, the settings they pick and the flow AI video shot they line up
A TikTok creator writes each hook as a one-line prompt, then sets Kling V3 Turbo to 9:16, 1080p and 10 seconds. Seven hooks, seven prompts, one ratio, so every flow AI video post that week keeps the same vertical frame.
An indie director drafts a 21:9 establishing shot on Seedance 2.0, pasting a full scene description that fits well inside its 20,000-character limit. For the handoff between two scenes, Veo's start and end frame mode takes the last still of one scene and the first of the next.
A marketer keeps one product prompt and changes only the frame: 16:9 for YouTube, 9:16 for Stories, 1:1 for the feed. HappyHorse 1.1 covers all three at 720p or 1080p, so each flow AI video variant differs by ratio, not by message.
A physics teacher adds a diagram as the starting image, then sets Wan 2.6 Image to Video to 1080p and 15 seconds for a flow AI video walkthrough. For slide decks, Wan 2.7 also lists a 4:3 frame.
A creator running a mascot account keeps one character sheet and adds it as the reference image in Veo Fast. Every flow AI video episode starts from that reference in 16:9, and the prompt, capped at 2,000 characters, only changes the scene.
Someone making a first flow AI video for a birthday picks a favorite photo, chooses Hailuo 2.3 Standard for image-only input, and sets 6 seconds at 1080P. The prompt allows 5,000 characters, yet one clear action line such as 'grandma waves and laughs' is a good start.
Straight answers on flow AI video models, prompts, phones and pricing
Yes. The flow AI video studio runs in your mobile browser, so there is nothing to install. Pick a model, type the prompt, and choose 9:16 for a vertical frame; Veo, Kling, Seedance, Wan 2.7, Grok and HappyHorse all list it. Finished clips download straight to your phone.
Choose a model for your flow AI video, write the prompt, then set ratio, duration and resolution. Veo Fast is preselected; switch to Seedance 2.0 for 21:9 or 4K, or to Hailuo 2.3 Pro when you begin with a photo. Durations run from 4 to 15 seconds depending on the model.
Match the model to the shot type. Seedance 2.0 suits wide 21:9 scenes and long prompts, up to 20,000 characters. Veo suits shots where the first and last frame matter, or where a reference image should guide the look. For a photo as the source, try Hailuo 2.3 or Grok Video 1.5 Preview; for 4K, Kling 3.0 or Seedance 2.0.
Write it as a shot description in five parts: subject, action, camera move, light and length. For example: 'a cyclist crests a hill at dawn, slow dolly left, warm backlight, 10 seconds.' Keep one action per flow AI video clip, and check the limit: 1,000 characters on Kling 2.6, 2,000 on Veo, 20,000 on Seedance.
You can start free: new accounts get free credits to try the flow AI video studio, and each clip uses credits based on the model, resolution and length you pick. When you need more, plans and credit packs are listed on the Pricing page.
Give the model fewer decisions to make. Use Veo's start and end frame mode to fix where motion begins and ends, or a reference image on Veo Fast and Veo Lite to hold a look in 16:9. Pick a shorter length, such as 5 seconds on Kling or Wan 2.6, and describe one action per prompt.
Write to us
Ask about models or settings
Replies by email
