---
title: "AI video models"
description: "Every AI video generation model on Hedra, grouped by the lab that makes it."
canonical: "https://www.hedra.com/models/video"
---

# AI video models

[View on Hedra](https://www.hedra.com/models/video)

## [Creatify Aurora](https://www.hedra.com/models/video/creatify/aurora.md)

Create high-fidelity speaking or singing avatar videos.

480p 7¢/second, 720p 14¢/second; a partial second bills whole.

## [FLUX.3](https://www.hedra.com/models/video/black-forest-labs/flux-3.md)

Set a still in motion, with native audio.

720p 17¢/second, 1080p 29¢/second.

## [FLUX.3 Video Upscaler Creative](https://www.hedra.com/models/video/black-forest-labs/flux-3-video-upscaler-creative.md)

Super-resolution that adds detail as it enlarges, for soft or low-resolution footage.

1080p 20¢/second, 2K 35¢/second, 4K 79¢/second.

## [FLUX.3 Video Upscaler Precise](https://www.hedra.com/models/video/black-forest-labs/flux-3-video-upscaler-precise.md)

Super-resolution that reproduces the source faithfully, for footage whose detail should not change.

1080p 14¢/second, 2K 25¢/second, 4K 55¢/second.

## [Gemini Omni 1.1 Flash](https://www.hedra.com/models/video/google/gemini-omni-flash-11.md)

Gemini's fast multimodal video model — cinematic clips with native audio, from a prompt, a keyframe pair, or reference images, up to 4K.

360p 3.55¢/second, 720p 10.64¢/second, 1080p 15.96¢/second, 4k 31.93¢/second; plus $1.575 per 1M input tokens: 1,100 per image, one per prompt byte (at least 4,096 when the prompt is enhanced) plus 1,200, and per frame at the clip's own frame rate, up to 240 a second, 95 for each reference video and 245 for an edit's source clip (18 and 48 at 360p), plus 25 per second of video input; and 1,000 text output tokens at $9.45 per 1M; an edit bills its output for the source clip's length, rounded up to the next sixth of a second, up to 10 seconds.

## [Gemini Omni 1.2 Flash](https://www.hedra.com/models/video/google/gemini-omni-flash-12.md)

Gemini's fast multimodal video model — cinematic clips of up to 15 seconds with native audio, from a prompt, a keyframe pair, or reference images, up to 4K.

360p 2.43¢/second, 720p 7.3¢/second, 1080p 10.95¢/second, 4k 21.89¢/second; plus $1.575 per 1M input tokens: 1,100 per image, one per prompt byte (at least 4,096 when the prompt is enhanced) plus 1,200, and per frame at the clip's own frame rate, up to 240 a second, 95 for each reference video and 245 for an edit's source clip (18 and 48 at 360p), plus 25 per second of video input; and 1,000 text output tokens at $9.45 per 1M; an edit bills its output for the source clip's length, rounded up to the next sixth of a second, up to 15 seconds.

## [Gemini Omni Flash](https://www.hedra.com/models/video/google/gemini-omni-flash.md)

Gemini's fast multimodal video model — cinematic clips with native audio from a prompt, a keyframe, or reference images.

720p 12.67¢/second; plus $1.875 per 1M input tokens: 1,100 per image, one per prompt byte (at least 4,096 when the prompt is enhanced) plus 1,200, and per frame at the clip's own frame rate, up to 240 a second, 95 for each reference video and 245 for an edit's source clip, plus 25 per second of video input; and 1,000 text output tokens at $11.25 per 1M; an edit bills its output for the source clip's length, rounded up to the next sixth of a second, up to 10 seconds.

## [Happy Horse](https://www.hedra.com/models/video/alibaba/happy-horse.md)

Open-weight video from a prompt, from a first frame, or from up to nine reference images you direct by name as character1..character9.

720p 21.43¢/second, 1080p 42.86¢/second.

## [HeyGen Photo Avatar 4](https://www.hedra.com/models/video/heygen/photo-avatar-4.md)

Turn a clear portrait and driving audio into a talking avatar.

360p/480p/540p/720p/1080p 10¢/second; a partial second bills whole.

## [Kling 1.6](https://www.hedra.com/models/video/kling/16.md)

Text: 720p 9.8¢/second; Image: 1080p 9.8¢/second.

## [Kling 2.1 Master](https://www.hedra.com/models/video/kling/21-master.md)

Cinema-grade video with striking textures and rich depth.

720p 28¢/second.

## [Kling 2.5 Turbo](https://www.hedra.com/models/video/kling/25-turbo.md)

Fast, high-quality video generation.

1080p 7¢/second.

## [Kling 2.6 Motion Control](https://www.hedra.com/models/video/kling/26-motion-control.md)

Transfer movements from a reference video to any character image. Cost-effective mode for motion transfer, perfect for portraits and simple animations.

720p 7¢/second; a partial second bills whole; 1080p 11.2¢/second; a partial second bills whole.

## [Kling 2.6 Pro](https://www.hedra.com/models/video/kling/26-pro.md)

Cinematic visuals, fluid motion, and native audio generation.

1080p 14¢/second with audio, 7¢/second without.

## [Kling AI Avatar v2](https://www.hedra.com/models/video/kling/ai-avatar-v2.md)

Create avatar videos with realistic humans, animals, cartoons, or stylized characters from an image and audio input.

Standard: 720p 5.62¢/second; a partial second bills whole; Pro: 720p 11.5¢/second; a partial second bills whole.

## [Kling O1](https://www.hedra.com/models/video/kling/o1.md)

Generate from a single image with text-driven style and scene guidance.

1080p 14.29¢/second.

## [Kling O3](https://www.hedra.com/models/video/kling/o3.md)

Clips up to 15 seconds with native audio.

Standard: 720p 11.2¢/second with audio, 8.4¢/second without; Pro: 1080p 14¢/second with audio, 11.2¢/second without, 4K 42¢/second.

## [Kling O3 Edit](https://www.hedra.com/models/video/kling/o3-edit.md)

Edit videos using natural language.

720p 12.6¢/second; billed for the source video's length, or for the nearest whole second when that is longer; 1080p 16.8¢/second; billed for the source video's length, or for the nearest whole second when that is longer; 4K 42¢/second; billed for the source video's length, or for the nearest whole second when that is longer.

## [Kling O3 Reference](https://www.hedra.com/models/video/kling/o3-reference.md)

Input a reference video and preserve motion and camera style.

720p 12.6¢/second; billed for the source video's length, and for at least 5 seconds; 1080p 16.8¢/second; billed for the source video's length, and for at least 5 seconds.

## [Kling V3](https://www.hedra.com/models/video/kling/v3.md)

Ultra-high-definition storyboards with native audio.

Standard: 720p 12.6¢/second with audio, 8.4¢/second without; Pro: 1080p 16.8¢/second with audio, 11.2¢/second without, 4K 42¢/second.

## [Kling V3 Motion Control](https://www.hedra.com/models/video/kling/v3-motion-control.md)

Animate a character image to match the motion of a reference video. Standard tier for cost-effective generation.

720p 12.6¢/second; a partial second bills whole; 1080p 16.8¢/second; a partial second bills whole.

## [LTX-2.3](https://www.hedra.com/models/video/lightricks/ltx-2-3.md)

Animate a start frame up to 4K, optionally toward an end frame, with native audio.

Fast: 1080p 6¢/second, 1440p 12¢/second, 2160p 24¢/second; Pro: 1080p 8¢/second, 1440p 16¢/second, 2160p 32¢/second.

## [LTX-2.5](https://www.hedra.com/models/video/lightricks/ltx-2-5.md)

Clips up to 20 seconds with synchronized native audio, from a prompt, a start frame, or a start and end frame, or cut to a supplied audio track — on the quick path.

Fast Text/Image: 720p 9¢/second, 1080p 13¢/second, 1440p 19¢/second, 2160p 30¢/second; Fast Audio: 1080p 13¢/second; Pro Text/Image: 720p 12¢/second, 1080p 17¢/second; Pro Audio: 1080p 17¢/second.

## [Luma Ray 3.2](https://www.hedra.com/models/video/luma-ai/luma-ray-32.md)

Cinematic motion with deliberate camera control, from a prompt.

540p 10¢/second, 720p 20¢/second, 1080p 40¢/second.

## [MiniMax H3](https://www.hedra.com/models/video/minimax/h3.md)

One model for every starting point — a prompt, a keyframe pair, or reference images that keep a subject consistent.

480p 5¢/second, 768p 6¢/second, 2K 13¢/second, 4K 16¢/second.

## [MiniMax H3 Max](https://www.hedra.com/models/video/minimax/h3-max.md)

MiniMax H3 Max video generation from text, keyframes, or reference assets.

Reference images: 480p/768p 8¢/second; no 480p discount in reference mode; plus reference inputs at $0.02/1,000 billable tokens; pool tokens across all inputs, then subtract 4,096 included tokens (minimum zero): each image contributes ceil(width × height / 1024), each video contributes ceil(seconds × 2,886) at 480p or ceil(seconds × 7,459.2) at 768p, each audio contributes ceil(seconds × 80); input dimensions and durations are required for a complete quote; Text/Image/Start/end images: 480p 5¢/second, 768p 8¢/second.

## [MiniMax H3 Max Camera Controls](https://www.hedra.com/models/video/minimax/h3-max-camera-controls.md)

Orbit, tilt, and dolly a camera through a single image along a path you key-frame in 3D.

480p 5¢/second, 768p 8¢/second.

## [MiniMax H3 Max Turbo](https://www.hedra.com/models/video/minimax/h3-max-turbo.md)

MiniMax H3 Max Turbo — the same text and keyframe inputs, served faster and at half the price.

480p 2.5¢/second, 768p 4¢/second.

## [MiniMax Hailuo 2.3](https://www.hedra.com/models/video/minimax/hailuo-23.md)

Everyday 1080p video with natural movement.

Standard: 28¢/6-second video, 56¢/10-second video; Pro: 49¢/6-second video; Fast Standard: 19¢/6-second video, 32¢/10-second video; Fast Pro: 33¢/6-second video.

## [MiniMax Hailuo-02](https://www.hedra.com/models/video/minimax/hailuo-02.md)

Everyday 1080p video with natural movement.

Standard: 768p 4.5¢/second; Pro: 1080p 8¢/second.

## [Omnihuman 1.5](https://www.hedra.com/models/video/bytedance/omnihuman-15.md)

Creates vivid, emotional character videos driven entirely by your audio.

720p/1080p 16¢/second; a partial second bills whole.

## [Omnihuman 2.0](https://www.hedra.com/models/video/bytedance/omnihuman-20.md)

Performs your audio's tone and emotion in one take of up to five minutes.

720p 10.5¢/second; a partial second bills whole.

## [PixVerse V6](https://www.hedra.com/models/video/pixverse/v6.md)

Animate a still into a stylized 1080p clip up to 15 seconds, with native audio.

360p 3.5¢/second with audio, 2.5¢/second without, 540p 4.5¢/second with audio, 3.5¢/second without, 720p 6¢/second with audio, 4.5¢/second without, 1080p 11.5¢/second with audio, 9¢/second without.

## [Seedance 2.0](https://www.hedra.com/models/video/bytedance/seedance-20.md)

Reference-driven video up to 4K with native audio — hold a look across shots with reference images, clips, or audio.

Standard: 4K $1.0968/second, 480p 16.07¢/second, 720p 21.43¢/second, 1080p 52.73¢/second; plus each second of reference video at 60% of the output's per-second rate; reference video durations are required for a complete quote; Fast: 480p 12.86¢/second, 720p 17.14¢/second; plus each second of reference video at 60% of the output's per-second rate; reference video durations are required for a complete quote; Mini: 480p 8.04¢/second, 720p 10.71¢/second; plus each second of reference video at 60% of the output's per-second rate; reference video durations are required for a complete quote.

## [Seedance 2.5](https://www.hedra.com/models/video/bytedance/seedance-25.md)

Reference-driven video up to 30 seconds at 1080p, with native audio.

480p 14.85¢/second, 720p 32.14¢/second, 1080p 79.09¢/second; plus each second of reference video at 60% of the output's per-second rate; reference video durations are required for a complete quote; an edit of the reference video, chosen from the prompt, costs the 30-second maximum, less a refund for the seconds it does not render.

## [Topaz Hyperion 2.5 HDR](https://www.hedra.com/models/video/topaz-labs/topaz-video-upscaler-hyperion-2-5.md)

Convert SDR video to 10-bit HDR with richer highlights, color, and tonal separation. The output keeps the source resolution.

by source size: long edge under 2160 px $11.0592/1 minute, long edge 2160 px or more $24.1655/1 minute.

## [Topaz Proteus](https://www.hedra.com/models/video/topaz-labs/topaz-video-upscaler.md)

Precision upscaling that cleans compression and noise while staying faithful to the source.

1080p 2.14¢/second, 2K/4K 8.57¢/second.

## [Topaz Starlight Fast](https://www.hedra.com/models/video/topaz-labs/topaz-video-upscaler-starlight-fast.md)

Faster generative diffusion upscaling at half the cost of Starlight Precise.

by output size: long edge under 2160 px 4.61¢/second, long edge 2160 px or more 10.07¢/second.

## [Topaz Starlight HQ](https://www.hedra.com/models/video/topaz-labs/topaz-video-upscaler-starlight-hq.md)

Generative diffusion upscaling balancing detail and sharpness for medium-to-high quality sources.

by output size: long edge under 2160 px 9.22¢/second, long edge 2160 px or more 20.14¢/second.

## [Topaz Starlight Precise](https://www.hedra.com/models/video/topaz-labs/topaz-video-upscaler-starlight-precise.md)

Generative diffusion upscaling for AI-generated and archival video with realistic faces, textures, and text.

by output size: long edge under 2160 px 9.22¢/second, long edge 2160 px or more 20.14¢/second.

## [VEED Fabric 1.0](https://www.hedra.com/models/video/veed/fabric-10.md)

Talking video with natural lip-sync and expressive animation.

480p 11.43¢/second, 720p 21.43¢/second; the charge rounds up to a whole multiple of the 480p per-second rate.

## [VEED Video Background Removal](https://www.hedra.com/models/video/veed/video-background-removal.md)

Remove a video's background and return transparent WebM.

2.25¢/30 frames; a partial block of 30 frames bills whole.

## [Veo 3](https://www.hedra.com/models/video/google/veo-3.md)

Hollywood-grade, cinematic video straight from text—your go-to for hero campaigns.

Standard: 720p/1080p 40¢/second with audio, 20¢/second without; Fast: 720p/1080p 15¢/second with audio, 10¢/second without.

## [Veo 3.1](https://www.hedra.com/models/video/google/veo-31.md)

For unparalleled detail and nuance, perfect for when your vision requires the best possible quality.

Standard: 720p/1080p 40¢/second with audio, 20¢/second without; Fast: 720p 14.29¢/second with audio, 11.43¢/second without, 1080p 16.46¢/second with audio, 13.71¢/second without.

## [Vidu Q3](https://www.hedra.com/models/video/vidu/q3.md)

The longest clips in the catalog — up to 16 seconds with native dialogue and sound, from a text prompt, from a start frame, or between a start and end frame

Standard: 540p 7¢/second, 720p/1080p 15.4¢/second; Turbo: 540p 3.5¢/second, 720p/1080p 7.7¢/second.

## [Vidu Q3 Reference](https://www.hedra.com/models/video/vidu/q3-reference.md)

Keep up to four subjects consistent across a clip from reference images.

540p 7¢/second, 720p/1080p 15.4¢/second.

## [Wan 2.7](https://www.hedra.com/models/video/alibaba/wan-2-7.md)

Wan 2.7 video with native audio — from a text prompt, from a first frame with an optional last frame, or from reference images that keep subjects consistent

Text/Image: 720p 10¢/second, 1080p 15¢/second; Reference images: 720p/1080p 10¢/second.

## [Wan 3.0](https://www.hedra.com/models/video/alibaba/wan-3-0.md)

Wan 3.0 video with native audio, up to 30 seconds in one shot — from a text prompt, from a first frame with an optional last frame, or from reference images that keep subjects consistent

Standard Reference images: 480p 5¢/second, 720p 10¢/second, 1080p 20¢/second; plus each second of reference video at the output's per-second rate; reference video durations are required for a complete quote; Standard Text/Image: 480p 5¢/second, 720p 10¢/second, 1080p 20¢/second; Prime Reference images: 480p 6.8¢/second, 720p 14¢/second, 1080p 28¢/second; plus each second of reference video at the output's per-second rate; reference video durations are required for a complete quote; Prime Text/Image: 480p 6.8¢/second, 720p 14¢/second, 1080p 28¢/second.

[Model API documentation](https://www.hedra.com/docs/pages/developer/v3/model-jobs.md)
