GPT Image 2.5 is live — OpenAI's newest image model, targeted edits that leave the rest of the frame alone
All resources

· reAPI Team

Vidu Q4 release date: what launched in the Preview

Vidu Q4 Preview launched on October 7, 2026. Check the official release record, image and audio inputs, duration limits, and what Preview means before using it.

Vidu Q4 release date: what launched in the Preview

The Vidu Q4 release date is October 7, 2026, for the Preview version. Vidu's official update notice lists that date next to the new model. Its model directory still labels Vidu Q4 as Preview, and its API documentation uses the request ID viduq4-preview.[1][2][3]

The next decision is practical: which inputs should you prepare for Vidu Q4? The documented choices are a single image used as the opening frame, or a set of references used to compose a scene. This guide explains the difference, the input limits and the meaning of the Preview label. The facts below were checked against Vidu's own pages on October 22, 2026; this is a documentation review, with no generation benchmark behind it.

The cover is a conceptual illustration, not a generated-video sample or a product screenshot.

TL;DR

  • Vidu's October 7 update introduces Vidu Q4 Preview. The announcement date applies to that version.[1]
  • The official model directory currently identifies Vidu Q4 as Preview. The sources reviewed do not announce a separate final-version release date.[2]
  • Image-to-video accepts one opening image. Reference-to-video accepts 1–15 images and up to three MP3 references.[3][4]
  • Both Vidu Q4 APIs document 3–16 seconds, with a five-second default, and list resolutions through 4K.[3][4]
  • Vidu describes native audio, reference voices and automatic shot changes.[1][4] For a voice-led scene, prepare a reference recording and a short line of dialogue.

The Vidu Q4 release date belongs to the Preview

Vidu records the date in its official update notice. Its October 7, 2026 entry introduces Vidu Q4 Preview and describes voice references, more reference images, motion and visual-effects changes, and 2K/4K output. The entry is part of Vidu's API update history, so it gives a dated record you can check directly.[1]

Two other official sources identify the same version. The model directory displays a Preview badge beside Vidu Q4. The image-to-video and reference-to-video documentation both name viduq4-preview in the request's model field.[2][3][4]

Put the full version name in your project brief or evaluation record. A note that says only "tested Q4" leaves out which release was involved. Save the model ID with the date so you can identify the version when revisiting a result.

For a final, non-Preview version, the reviewed update log and current model directory provide no separate release date. October 7 should therefore remain attached to Preview. Check that log for a separate final-version announcement.

Choose the image mode before preparing the references

Vidu Q4's two modes use images differently. The distinction matters even if you have only one picture. Image-to-video uses it as the first frame, while reference-to-video uses its contents as a reference for the generated subject or scene.[3][4]

Your starting materialDocumented modeWhat to prepare
A finished still whose opening composition you want to animateImage-to-videoOne image, plus an optional motion description
Separate pictures of a character, object or locationReference-to-video1–15 reference images and a required prompt
A scene that also needs reference voicesReference-to-videoImage references, a prompt and up to three supported audio files

Consider a hypothetical product clip. If you already have a photograph with the product in the intended setting and want that photograph to start moving, image-to-video is the direct fit. If the product and setting are in separate pictures, reference-to-video gives you a way to provide both. Choose the mode that matches the role of your source pictures.

The official product page says single-image output inherits the image's aspect ratio. Reference mode instead exposes a choice of 16:9, 9:16, 1:1, 4:3 or 3:4.[5][4] Check that distinction before you prepare a vertical campaign from landscape source material.

One image can also be a valid reference-mode input. File count alone does not determine which mode you intended. In an application, make the mode choice explicit and retain it with the submitted references so a later revision uses the same interpretation.

The limits that affect your first Vidu Q4 clip

The current official API pages document the following settings. These describe Vidu's direct API; the interface and field names in any other service need their own documentation check.[3][4]

SettingImage-to-videoReference-to-video
ImagesExactly one opening image1–15 reference images
PromptOptional; up to 20,000 charactersRequired; up to 20,000 characters
Audio referencesNo reference-audio field documentedUp to three MP3 files
Output duration3–16 seconds; default 53–16 seconds; default 5
Resolution540p, 720p, 1080p, 2K, 4K540p, 720p, 1080p, 2K, 4K
Default resolution720p720p
Generated audioOn by default; can be disabledOn by default; can be disabled

For the single-image endpoint, Vidu lists PNG, JPEG, JPG and WebP, with a 50 MB image limit.[3] Prepare images for reAPI as publicly accessible HTTP(S) URLs. Its request guide explains the accepted media URL format and how to distinguish a first frame from a reference image.[6]

Keep the image URLs accessible while the task runs. If you add audio references, prepare MP3 files between 3 and 12 seconds long, with each file no larger than 50 MB. Reference mode accepts up to three of them.[4]

A longer written prompt cannot extend the documented 16-second output limit. For a piece longer than that, plan separate clips and an editing step. A proposed 30-second sequence could use two 15-second shots. Plan what happens at the cut, then inspect whether subjects and sound remain coherent across it.

Reference voices and generated sound serve different jobs

Vidu Q4 reference mode has separate controls for supplied audio references and generated audio. The supplied MP3 files guide the sound, while the audio switch controls whether the output includes generated audio. Generated audio is enabled by default.[4]

That separation helps when planning a scene. A voice sample is an input intended to guide the voice. The completed video's sound is an output to inspect. The official release describes greater voice consistency and more natural emotion.[1] Listen to the result alongside your reference to assess the voice for your scene.

For a proposed dialogue test, use a short line whose words you can check. Save the intended speaker and reference-file order alongside the text. Then review the entire clip for the requested words, speaker assignment and timing. Write down any missing words or changes of speaker before revising the prompt.

When a demonstration includes music, check its production notes before trying to recreate the soundtrack. The finished reel alone does not tell you whether its music came from the video model, another tool or an editing step.

What the launch record can tell you before a trial

The release notice describes improvements in performance, motion, effects and camera language.[1] Use those areas to choose what to inspect in a trial: a moving subject, a camera change, an effect or a spoken exchange.

A useful first trial needs an observable requirement. For an opening-image animation, that might be the subject staying recognizable while a requested camera movement occurs. For a reference-based scene, it might be whether each supplied subject appears in the intended role. Write the requirement before you generate, so a visually attractive result does not hide a missed instruction.

Keep the requested duration and resolution in the same record as the inputs. If you change the source image, prompt and settings together, the next output will be hard to interpret. Change one part at a time and compare the same moment in each version. Keep the earlier take until you decide whether the revision helped.

For reAPI-specific settings, use the Vidu Q4 model guide. Check the available controls and current billing information there before submitting a job.

Frequently asked questions

When was Vidu Q4 released?

Vidu's official update notice dates the Q4 Preview release to October 7, 2026. Keep "Preview" attached to that date when describing the release.[1]

Has the final version been released?

The official directory reviewed on October 22 still identifies Vidu Q4 as Preview. Neither that page nor the reviewed update log gives a separate final-version release date.[2][1]

Can I generate from a text prompt alone?

The current Vidu Q4 Preview documentation provides image-to-video and reference-to-video, both requiring image input. It does not document a Vidu Q4 text-only endpoint.[3][4]

Can I use one image in reference mode?

Yes. The documented reference range starts at one image. That image guides the subject; image-to-video is the separate mode that explicitly uses an image as the first frame.[4][3]

How can I make longer videos?

The documented Vidu Q4 APIs stop at 16 seconds per generation. For a longer piece, plan and edit several clips. Matching the subjects, cuts and audio across those clips remains a separate production task.[3][4]

Does the Preview support 4K and audio together?

Both official API pages list 4K among the resolution options and expose the audio switch, without a resolution-specific audio exclusion. Set the resolution and audio fields explicitly when you need both.[3][4]

Keep the version with your project notes

Record viduq4-preview, the chosen input mode and the documentation date before preparing a trial. The model guide is the place to check the reAPI-facing controls. If you return later to check the Vidu Q4 release date, open the official update log and look for a separate final-version announcement before changing the version in your notes.

References

  1. Vidu. Update notice: October 7, 2026, Vidu Q4 preview. Retrieved October 22, 2026. platform.vidu.com/docs/api-reference/updates.
  2. Vidu. Vidu models map. Retrieved October 22, 2026. platform.vidu.com/models.
  3. Vidu. Q4 Preview: image to video. Retrieved October 22, 2026. platform.vidu.com/docs/api-reference/video-models/vidu-q4-preview/image-to-video.
  4. Vidu. Q4 Preview: reference to video. Retrieved October 22, 2026. platform.vidu.com/docs/api-reference/video-models/vidu-q4-preview/reference-to-video.
  5. Vidu. Vidu Q4 product page. Retrieved October 22, 2026, Chinese edition. vidu.com/vidu-q4.
  6. reAPI. Vidu Q4 request guide. Request documentation reviewed October 22, 2026. reapi.ai/docs/vidu-q4.
All resources