2025 releases · AI video

    Every AI video model released in 2025

    70 models from 16 providers, dated 15 January 2025 to 16 December 2025. They did not arrive evenly — they landed in 13 bursts, and this page is those bursts in order.

    The 2025 video timeline

    15 January – 25 February 2025

    4 models · 3 providers

    Luma, Pixverse and Wan shipped 4 video models between 15 January 2025 and 25 February 2025. Of the 4, 3 have a spec page and 1 is folded into a parent model's page as tier or mode variants. Cheapest complete job in the window: 10 credits on Wan 2.1 Image to Video. Highest output in it: 1080p on Pixverse V4.

    • First image to video model in the catalog — Wan 2.1 Image to Video.
    • First text to video model in the catalog — Luma Ray 2 720p.
    • First 1080p output in the catalog — Pixverse V4.
    • Longest single clip the catalog had offered — 15s on Wan 2.1 Image to Video.
    ModelBuilt byCredits per jobMax outputClip lengths
    Luma Ray 2 720pLuma Ray 2 video generation in 720p qualityLuma50-100 credits720p5s, 9s
    Luma Ray Flash 2 720pvariantFast Luma Ray 2 video generation in 720pLuma20-40 credits720p5s, 9s
    Pixverse V4Pixverse version 4 for creative video generationPixverse15-40 credits1080p5s, 8s
    Wan 2.1 Image to VideoWan 2.1 is an open-source AI video generation model that utilizes a diffusion transformer architecture and a novel 3D…Wan10-50 credits1080p5s, 10s, 15s

    31 March – 20 May 2025

    3 models · 2 providers

    Google and Runway shipped 3 video models between 31 March 2025 and 20 May 2025. Of the 3, 1 has a spec page and 2 are folded into a parent model's page as tier or mode variants. Cheapest complete job in the window: 90 credits on Runway Image to Video. Highest output in it: 1080p on Runway Image to Video.

    • First model to generate its own audio track — VEO 3 Fast.
    • New names on the roster: Runway, Google.
    ModelBuilt byCredits per jobMax outputClip lengths
    Runway Image to VideovariantRunway 4 for image to videoRunway90-180 credits1080p5s, 10s
    Runway Text to VideoRunway 4 for text to video generationsRunway90-180 credits1080p5s, 10s
    VEO 3 FastvariantFast variant of Google VEO 3 video modelGoogle100-320 credits1080p4s, 6s, 8s

    29 May – 1 June 2025

    3 models · 3 providers

    Kling, Midjourney and Other shipped 3 video models between 29 May 2025 and 1 June 2025. Of the 3, 2 have a spec page and 1 has no page of its own. Cheapest complete job in the window: 35 credits on Midjourney V7 Image to Video. Highest output in it: 720p on Midjourney V7 Image to Video.

    • New names on the roster: Kling, Midjourney.
    ModelBuilt byCredits per jobMax outputClip lengths
    Kling V2.1Kling version 2.1 for advanced video generationKling70-140 credits5s, 10s
    Midjourney V7 Image to VideoMidjourney V7 Image to Video generates high-quality videos from input images with cinematic motion and style consistencyMidjourney35-110 credits720p
    Video Enhancer ProProfessional video enhancement and upscaling modelOther104-2484 credits

    11 June 2025

    6 models · 1 provider

    ByteDance shipped 6 video models on 11 June 2025. Of the 6, 2 have a spec page and 4 are folded into a parent model's page as tier or mode variants. Cheapest complete job in the window: 2 credits on ByteDance Seedance 1 Pro Fast. Highest output in it: 1080p on ByteDance Seedance 1 Lite.

    • New name on the roster: ByteDance.
    ModelBuilt byCredits per jobMax outputClip lengths
    ByteDance Seedance 1 LitevariantLightweight video generation for quick resultsByteDance7-105 credits1080p4s, 5s, 6s, 7s, 8s, 9s
    ByteDance Seedance 1 ProvariantProfessional video generation with high qualityByteDance5-146 credits1080p2s, 3s, 4s, 5s, 6s, 7s
    ByteDance Seedance 1 Pro FastvariantFast professional video generation from ByteDanceByteDance2-59 credits1080p2s, 3s, 4s, 5s, 6s, 7s
    Seedance Image to VideoTransform images into dynamic dancing videosByteDance4-105 credits1080p2s, 3s, 4s, 5s, 6s, 7s
    Seedance Text to VideovariantCreate dancing videos from text descriptionsByteDance4-105 credits1080p2s, 3s, 4s, 5s, 6s, 7s
    Seedance 1 Pro FastByteDance Seedance 1 Pro Fast - accelerated image-to-video generation with high visual qualityByteDance62-124 credits1080p5s, 10s

    18 June – 28 July 2025

    7 models · 3 providers

    Grok, Hailuo and Wan shipped 7 video models between 18 June 2025 and 28 July 2025. Of the 7, 3 have a spec page, 3 are folded into a parent model's page as tier or mode variants and 1 has no page of its own. Cheapest complete job in the window: 16 credits on Grok Imagine Extend. Highest output in it: 1080p on Hailuo 02.

    • First extend video model in the catalog — Grok Imagine Extend.
    • First motion control model in the catalog — Wan 2.2 14B Move.
    • Longest single clip the catalog had offered — 30s on Grok Imagine Image to Video.
    • New names on the roster: Hailuo, Grok.
    ModelBuilt byCredits per jobMax outputClip lengths
    Hailuo 02Creative video generation with artistic flairHailuo40-960 credits1080p
    Grok Imagine ExtendGrok Imagine video extension - continue an existing Grok Imagine video at a chosen timestamp, 6 or 10 secondsGrok16-80 credits2s, 3s, 4s, 5s, 6s, 7s
    Grok Imagine Image to VideovariantGrok Imagine Image to Video generates high-quality videos from input images with creative motion modes, multi-image support up to…Grok30-210 credits1080p6s, 10s, 15s, 20s, 30s
    Grok Imagine VideoGrok image-to-video generationGrok30-210 credits1080p6s, 10s, 15s, 20s, 30s
    Wan 2.2 14B MoveAdvanced video motion control with 14B parametersWan20-960 credits720p
    Wan 2.2 14B ReplacevariantIntelligent video object replacementWan20-960 credits720p
    Wan 2.2 Image to VideovariantWan 2.2 is an open-source AI video generation model that utilizes a diffusion transformer architecture and a novel 3D…Wan20-80 credits720p5s, 10s

    29 July – 1 September 2025

    4 models · 2 providers

    PrunaAI and VEED shipped 4 video models between 29 July 2025 and 1 September 2025. Of the 4, 1 has a spec page and 3 have no page of their own. Cheapest complete job in the window: 2 credits on PrunaAI P-Video. Highest output in it: 1080p on PrunaAI P-Video.

    • New names on the roster: PrunaAI, VEED.
    ModelBuilt byCredits per jobMax outputClip lengths
    PrunaAI P-VideoFast video generation with built-in draft mode for rapid creative iteration. Text-to-video, image-to-video, and audio-to-video in…PrunaAI2-40 credits1080p1s, 2s, 3s, 4s, 5s, 6s
    VEED Video Background RemovalRemove video backgrounds with AIVEED12-270 credits
    VEED Video Background Removal FastFast video background removal with AIVEED6-144 credits
    VEED Video Background Removal Green ScreenRemove video backgrounds and replace with green screenVEED13-300 credits

    18–24 September 2025

    6 models · 3 providers

    Decart, Kling and Wan shipped 6 video models between 18 September 2025 and 24 September 2025. Of the 6, 4 have a spec page and 2 are folded into a parent model's page as tier or mode variants. Cheapest complete job in the window: 5 credits on Lucy Restyle. Highest output in it: 1080p on Kling 2.5 Turbo.

    • New name on the roster: Decart.
    ModelBuilt byCredits per jobMax outputClip lengths
    Lucy RestyleAI-powered video restyling and transformationDecart5-120 credits720p
    Kling 2.5 TurboFast video generation with good quality balanceKling21-42 credits1080p5s, 10s
    Wan Video 2.5 Image to VideoHigh-quality image-to-video with Wan 2.5Wan25-150 credits1080p5s, 10s
    Wan Video 2.5 Image to Video FastvariantFast image-to-video generation with Wan 2.5Wan50-150 credits1080p5s, 10s
    Wan Video 2.5 Text to VideoHigh-quality text-to-video with Wan 2.5Wan25-150 credits1080p5s, 10s
    Wan Video 2.5 Text to Video FastvariantFast text-to-video generation with Wan 2.5Wan25-150 credits1080p5s, 10s

    30 September 2025

    5 models · 1 provider

    OpenAI shipped 5 video models on 30 September 2025. Of the 5, 2 have a spec page and 3 are folded into a parent model's page as tier or mode variants. Cheapest complete job in the window: 40 credits on Sora 2 Image to Video. Highest output in it: 1080p on Sora 2 Image to Video.

    • New name on the roster: OpenAI.
    ModelBuilt byCredits per jobMax outputClip lengths
    Sora 2 Image to VideovariantConvert images to videos with Sora 2OpenAI40-200 credits1080p4s, 8s, 12s, 16s, 20s
    Sora 2 Image to Video ProvariantProfessional image-to-video with Sora 2 ProOpenAI120-1400 credits1080p4s, 8s, 12s, 16s, 20s
    Sora 2 Pro StoryboardSora 2 with advanced storyboard scene generationOpenAI300-750 credits10s, 15s, 25s
    Sora 2 Text to VideoText-to-video generation with Sora 2OpenAI40-200 credits720p4s, 8s, 12s, 16s, 20s
    Sora 2 Text to Video ProvariantProfessional text-to-video with Sora 2 ProOpenAI120-1400 credits1080p4s, 8s, 12s, 16s, 20s

    15 October 2025

    7 models · 1 provider

    Google shipped 7 video models on 15 October 2025. Of the 7, 4 have a spec page and 3 are folded into a parent model's page as tier or mode variants. Cheapest complete job in the window: 20 credits on VEO 3.1 Fast. Highest output in it: 4K on VEO 3.1.

    • First reference to video model in the catalog — VEO 3.1 Reference to Video.
    • First 4K output in the catalog — VEO 3.1.
    ModelBuilt byCredits per jobMax outputClip lengths
    VEO 3.1Enhanced VEO 3 with improved quality and featuresGoogle80-320 credits4K4s, 6s, 8s
    VEO 3.1 FastvariantFast variant of VEO 3.1 for quick video generationGoogle20-60 credits4K4s, 6s, 8s
    VEO 3.1 Reference to VideoGoogle VEO 3.1 reference-to-video — generate an 8s cinematic clip whose subjects match up to 3 reference images, with native…Google160-320 credits4K8s
    VEO First Last FrameGenerate videos between first and last framesGoogle80-320 credits4K4s, 6s, 8s
    VEO First Last Frame FastvariantFast frame-to-frame video generationGoogle40-120 credits4K4s, 6s, 8s
    VEO 3.1 Extend VideoExtend video duration with VEO 3.1Google140-280 credits1080p7s
    VEO 3.1 Fast Extend VideovariantFast video extension with VEO 3.1Google70-105 credits1080p7s

    28 October 2025

    3 models · 1 provider

    Hailuo shipped 3 video models on 28 October 2025. All 3 have a spec page. Cheapest complete job in the window: 14 credits on Hailuo 2.3 Fast. Highest output in it: 1080p on Hailuo 2.3 Pro.

    ModelBuilt byCredits per jobMax outputClip lengths
    Hailuo 2.3 FastFast MiniMax Hailuo 2.3 image-to-video generation — quick, cost-efficient cinematic clips from a single image.Hailuo14-28 credits6s, 10s
    Hailuo 2.3 ProProfessional-grade video generation with enhanced featuresHailuo49-98 credits1080p6s, 10s
    Hailuo 2.3 StandardStandard video generation with good qualityHailuo28-56 credits1080p6s, 10s

    4 November – 1 December 2025

    11 models · 4 providers

    ByteDance, Kling, Runway and Sam shipped 11 video models between 4 November 2025 and 1 December 2025. Of the 11, 4 have a spec page and 7 are folded into a parent model's page as tier or mode variants. Cheapest complete job in the window: 4 credits on ByteDance Upscaler Video. Highest output in it: 4K on ByteDance Upscaler Video.

    • New name on the roster: Sam.
    ModelBuilt byCredits per jobMax outputClip lengths
    ByteDance Upscaler VideoAI-powered video upscaling from ByteDanceByteDance4-346 credits4K
    Sam 3 Video SegmentIntelligent video segmentation for object detection and precise video editingSam4-90 credits
    Kling O1 Image to VideoKling O1 image-to-video generation.Kling56-112 credits3s, 4s, 5s, 6s, 7s, 8s
    Kling O1 Reference to VideovariantKling O1 reference-to-video generation.Kling28-56 credits3s, 4s, 5s, 6s, 7s, 8s
    Kling O1 Standard Image to VideovariantKling O1 Standard image-to-video generation.Kling42-84 credits3s, 4s, 5s, 6s, 7s, 8s
    Kling O1 Standard Reference to VideovariantKling O1 Standard reference-to-video generation.Kling42-84 credits3s, 4s, 5s, 6s, 7s, 8s
    Kling O1 Standard Video to Video EditvariantKling O1 Standard video-to-video edit.Kling63-1512 credits
    Kling O1 Standard Video to Video ReferencevariantKling O1 Standard video-to-video reference.Kling63-126 credits3s, 4s, 5s, 6s, 7s, 8s
    Kling O1 Video to Video EditvariantKling O1 video-to-video edit.Kling84-2016 credits
    Kling O1 Video to Video ReferencevariantKling O1 video-to-video reference.Kling84-168 credits3s, 4s, 5s, 6s, 7s, 8s
    Runway Gen-4.5Runway Gen-4.5 video generation with state-of-the-art quality, cinematic motion, and precise camera controlsRunway250-500 credits5s, 10s

    3 December 2025

    4 models · 1 provider

    Kling shipped 4 video models on 3 December 2025. Of the 4, 1 has a spec page and 3 are folded into a parent model's page as tier or mode variants. Cheapest complete job in the window: 18 credits on Kling Video V2.6 Standard Motion Control. Highest output in it: 1080p on Kling Video V2.6 Pro Motion Control.

    ModelBuilt byCredits per jobMax outputClip lengths
    Kling Video V2.6 Pro Image to VideovariantProfessional image-to-video with Kling V2.6 ProKling70-140 credits5s, 10s
    Kling Video V2.6 Pro Motion ControlvariantProfessional motion control with Kling V2.6 ProKling56-1344 credits1080p
    Kling Video V2.6 Pro Text to VideoProfessional text-to-video with Kling V2.6 ProKling70-140 credits1080p5s, 10s
    Kling Video V2.6 Standard Motion ControlvariantStandard motion control with Kling V2.6Kling18-420 credits720p

    16 December 2025

    7 models · 2 providers

    ByteDance and Wan shipped 7 video models on 16 December 2025. Of the 7, 4 have a spec page and 3 are folded into a parent model's page as tier or mode variants. Cheapest complete job in the window: 3 credits on Seedance 1.5 Pro Image to Video. Highest output in it: 4K on Wan V2.6 Reference to Video.

    ModelBuilt byCredits per jobMax outputClip lengths
    Seedance 1.5 Pro Image to VideovariantProfessional image-to-video with Seedance V1.5 ProByteDance3-35 credits1080p4s, 5s, 6s, 7s, 8s, 9s
    Seedance 1.5 Pro Text to VideoProfessional text-to-video with Seedance V1.5 ProByteDance5-70 credits1080p4s, 5s, 6s, 7s, 8s, 9s
    Wan 2.6 Image to VideoAlibaba WAN 2.6 is a state-of-the-art image-to-video generation model with audio support, prompt expansion, and multiple…Wan25-113 credits1080p5s, 10s, 15s
    Wan 2.6 Image to Video FlashUltra-fast image-to-video conversion with Wan 2.6 Flash technologyWan25-113 credits1080p5s, 10s, 15s
    Wan V2.6 Image to VideovariantAlibaba WAN 2.6 is a state-of-the-art image-to-video generation model with audio support, prompt expansion, and multiple…Wan50-225 credits1080p5s, 10s, 15s
    Wan V2.6 Reference to VideovariantGenerate videos from reference images with WAN V2.6Wan50-150 credits4K5s, 10s
    Wan V2.6 Text to VideoText-to-video generation with WAN V2.6Wan50-225 credits1080p5s, 10s, 15s

    When 2025 was busy

    10 of the twelve months carried a video release; the busiest window was 4 November – 1 December 2025, with 11.

    January 2025
    2
    February 2025
    2
    March 2025
    2
    May 2025
    2
    June 2025
    9
    July 2025
    7
    September 2025
    14
    October 2025
    10
    November 2025
    2
    December 2025
    20

    Who shipped video in 2025

    Kling (14), Wan (13) and ByteDance (9) led on volume. SKUs, not quality — four tiers of one model count four times.

    ProviderModelsSpec pagesFirstLatest
    Kling14429 May 20253 December 2025
    Wan13725 February 202516 December 2025
    ByteDance9411 June 202516 December 2025
    Google8420 May 202515 October 2025
    OpenAI5230 September 202530 September 2025
    Hailuo4318 June 202528 October 2025
    Grok3228 July 202528 July 2025
    Runway3231 March 20251 December 2025
    VEED301 September 20251 September 2025
    Luma2115 January 202515 January 2025
    Decart1118 September 202518 September 2025
    Midjourney111 June 20251 June 2025
    Other101 June 20251 June 2025
    Pixverse1125 February 202525 February 2025
    PrunaAI1129 July 202529 July 2025
    Sam1119 November 202519 November 2025

    What 2025 moved

    Firsts, measured against every video model released before them. “New name on the roster” is the provider label in the catalog, not the company — one lab can hold several labels.

    • 15 January – 25 February 2025
      • First image to video model in the catalog — Wan 2.1 Image to Video.
      • First text to video model in the catalog — Luma Ray 2 720p.
      • First 1080p output in the catalog — Pixverse V4.
      • Longest single clip the catalog had offered — 15s on Wan 2.1 Image to Video.
    • 31 March – 20 May 2025
      • First model to generate its own audio track — VEO 3 Fast.
      • New names on the roster: Runway, Google.
    • 29 May – 1 June 2025
      • New names on the roster: Kling, Midjourney.
    • 11 June 2025
      • New name on the roster: ByteDance.
    • 18 June – 28 July 2025
      • First extend video model in the catalog — Grok Imagine Extend.
      • First motion control model in the catalog — Wan 2.2 14B Move.
      • Longest single clip the catalog had offered — 30s on Grok Imagine Image to Video.
      • New names on the roster: Hailuo, Grok.
    • 29 July – 1 September 2025
      • New names on the roster: PrunaAI, VEED.
    • 18–24 September 2025
      • New name on the roster: Decart.
    • 30 September 2025
      • New name on the roster: OpenAI.
    • 15 October 2025
      • First reference to video model in the catalog — VEO 3.1 Reference to Video.
      • First 4K output in the catalog — VEO 3.1.
    • 4 November – 1 December 2025
      • New name on the roster: Sam.

    Dates are each model’s released_at value, walked in order and grouped until a window held 3 or more. Credits are what one complete generation costs per the model’s own price matrix; where it states none, the headline rate is shown and the model sits out the cheapest-in-window line.

    Other release years

    Frequently asked questions

    How many AI video models were released in 2025?+

    Versely's catalog carries 70 video models with a 2025 release date, from 16 providers, arriving in 13 launch windows between 15 January 2025 and 16 December 2025.

    What was the biggest AI video launch of 2025?+

    4 November – 1 December 2025, with 11 video models from ByteDance, Kling, Runway and Sam. Of the 11, 4 have a spec page and 7 are folded into a parent model's page as tier or mode variants. Cheapest complete job in the window: 4 credits on ByteDance Upscaler Video. Highest output in it: 4K on ByteDance Upscaler Video.

    Which company released the most AI video models in 2025?+

    Kling, with 14 of the 70 video models dated 2025 — first on 29 May 2025, most recently on 3 December 2025. Wan shipped 13, ByteDance shipped 9, Google shipped 8.

    What changed in AI video generation in 2025?+

    Measured against everything the catalog carried before it: First image to video model in the catalog — Wan 2.1 Image to Video; First text to video model in the catalog — Luma Ray 2 720p; First 1080p output in the catalog — Pixverse V4.

    Run any 2025 video model in Versely

    The all-in-one AI studio for creators. 60+ models for video, image, voice, music and lipsync in a single app.