PixVerse V4.5 Text to Video
PixVerse's May 2025 model and its last 4.x release, turning text into clips of 5 or 8 seconds at resolutions up to 1080p. PixVerse launched it promising smoother handling of complex movement.
Overview
PixVerse V4.5 came out on 13 May 2025 and was the last 4.x model before V5, which PixVerse's API changelog dates to 29 August 2025. The launch post named more than 20 cinematic camera controls, multi-element reference with Fusion, and smoother handling of complex movements. This page runs V4.5 as plain text-to-video, so only the last of those applies.
Choosing a variant: this page has the newest model of the three, which makes it the default for scenes with a lot of movement. PixVerse V4 Text to Video, from February 2025, is mainly worth keeping for projects that already have V4 footage to match. PixVerse V3.5 Text to Video is the oldest, from December 2024, and PixVerse highlighted anime when it launched.
What you can build
- Nature and wildlife footage, like the sample of a deer walking through a misty forest crossed by shafts of light.
- Night city and sci-fi scenes. One sample, made with the cyberpunk preset, shows cloaked figures on an illustrated street lined with glowing signs, with steam drifting across it.
- People moving through a scene, as in the sample of a toddler running at the camera through tall grass.
- Openers for ads and social posts in 16:9, 1:1 or 9:16.
Inputs
What to generate: a required prompt describing the scene, plus an optional negative prompt for anything you want left out.
Frame and length: aspect ratio is 16:9 unless you switch to 4:3, 1:1, 3:4 or 9:16. Duration is 5 seconds unless you choose 8.
Resolution and motion: quality starts at 540p and can be set to 360p (Turbo), 720p or 1080p. Motion mode is normal by default; fast is available for 5-second clips below 1080p, and other combinations run in normal mode.
Extras: a style menu with anime, 3d_animation, clay, comic and cyberpunk presets, and a seed from 0 to 2147483647 (empty means random).
Outputs
The model returns one MP4 video without sound. Three of the four samples on this page measure 1280x720 and the fourth 1024x576. All are 16:9 and roughly 5.4 seconds long.
Recommended settings
- Give the prompt room. PixVerse's guide puts the useful length at 25 to 200 words, with the motion described as well as the subject.
- Keep the seed fixed while you compare prompt edits, and settle the resolution before you start, since the same prompt and seed at a different quality give a different clip.
Limitations
- At 1080p a clip can only be 5 seconds long, in normal motion mode.
- The camera presets and Fusion that PixVerse announced with V4.5 are not settings on this page. Describe camera moves in the prompt instead, for example "the camera slowly pulls back to reveal the skyline".
- Input is text only, with no image field, and the video has no sound.
Safety & compliance
PixVerse screens V4.5 requests with content moderation, and a flagged request ends without a video. The PixVerse Platform Terms of Service govern use. PixVerse notes that subscribing grants no rights in third-party IP, a real person's likeness, music or brand assets; clear those rights before you prompt for them.
Example prompts
Great starting points for text-to-video-v4.5.
API quick start
Run text-to-video-v4.5 with a single API call.
{
"prompt": "A dynamic, storybook-style illustration o…",
"negativePrompt": "...",
"aspectRatio": "16:9",
"duration": 5
}curl -X POST "https://api.wiro.ai/v1/Run/pixverse/text-to-video-v4-5" \
-H "Content-Type: application/json" \
-H "x-api-key: YOUR_WIRO_API_KEY" \
--data-binary @- <<'JSON'
{
"prompt": "A dynamic, storybook-style illustration o…",
"negativePrompt": "...",
"aspectRatio": "16:9",
"duration": 5
}
JSONOther models you might like
Explore similar video generation models.