AI&Design
All archives
FEBRUARY 2025 — NOVEMBER 2026 · KST

Wan

Compare Wan video generation, audio, references, and editing from open releases to Alibaba Cloud, Runway, and Higgsfield. Track the Standard and Prime routes separately.

Posts archived
Attachments examined
Capabilities with evidence

What changed, when

Announcement dates in KST · Select a change or open its example.

Result example attached Announcement or contextual mediaBlank lane · no qualifying announcement in this source

Workflow threads

Read the dated updates and sources
  1. · Wan 2.1 releases inference code and model weights

    The open release offers routes for videos from text and still images. This is the date recorded by the Wan project, separate from later hosted API releases.

  2. · Wan 2.1 adds first- and last-frame generation

    The team publishes FLF2V weights and code to generate motion between two endpoint images. FLF2V is a named open Wan 2.1 model. Hosted product limits can differ.

  3. · VACE adds video creation and editing routes

    The project publishes VACE code, weights and a report for video editing, reference-driven generation, and extending or reframing footage. The examples are research demonstrations from the VACE project site.

  4. · Wan 2.2 releases open text and image routes

    The team ships Wan 2.2 text-to-video and image-to-video packages plus a combined text/image-to-video model. The team identifies these as open code and model-weight releases. ComfyUI and Diffusers support listed in the project record is separate from API hosting.

  5. · Wan 2.2 S2V adds audio-driven video

    Provide an image and audio to generate a performance synchronized with the input audio. The audio is an input to this named S2V model. This is separate from later routes that generate audio from text.

  6. · Wan 2.5 Preview adds synchronized generated audio

    The hosted text-to-video and image-to-video preview generates synced audio and video for clips up to ten seconds. The date is from Alibaba Cloud's International release section. The mainland China section records a separate September 19 listing.

  7. · Wan 2.6 adds longer multi-shot narrative

    Alibaba Cloud lists Wan 2.6 text-to-video and image-to-video routes for multi-shot scenes, up to 15 seconds, with voice generation in the image route. The date identifies the Model Studio text and image routes. Reference-to-video has its own release entry.

  8. · Wan 2.6 adds a reference-to-video route

    The reference endpoint can guide the appearance and voice of a specified subject and support multiple performers. Alibaba Cloud identifies the endpoint as wan2.6-r2v.

  9. · Alibaba lists a faster Wan 2.6 reference variant

    Wan 2.6 R2V Flash brings a separate, faster reference-to-video option to Model Studio. The release ledger names wan2.6-r2v-flash. It is distinct from later Wan 3.0 Prime.

  10. · Wan 2.7 adds a dedicated video-editing endpoint

    The public release notes list local or global video edits from prompts and image references, alongside text, image, and reference generation. The demos linked here come from separately published API guides. Their attachment documents the feature, while the release ledger supplies the chart date.

  11. · Higgsfield adds Wan 2.7 with frame control

    The Video tool introduces 1080p generation with endpoint frame control. The displayed April 6 notice timestamp converts to April 7 in KST. The entry covers Higgsfield's surface.

  12. · Wan 3.0 Standard reaches 30 seconds

    Model Studio combines generation, multimodal references, edits, and extension. It accepts text, images, video and audio, with document or public webpage input on applicable routes. Model Studio documents 480p–1080p and native audio. Input video plus output duration is capped at 30 seconds in the documented extension workflow.

  13. · Alibaba adds higher-speed Wan 3.0 Prime

    Prime is documented as a faster Wan 3.0 route with multimodal references and clips up to 30 seconds. Model Studio lists the Prime endpoint separately as wan3.0-video-prime.

  14. · Higgsfield adds Wan 3.0 and Prime

    Both variants are listed for video from a prompt or image up to 30 seconds and 1080p, with synchronized audio and multiple subjects. Higgsfield's product limits apply here. Alibaba's document-input capability is recorded separately.

  15. · Runway adds Wan 3.0 to Tool Mode and Workflows

    The Runway app changelog lists Wan 3.0 for generation in both surfaces. The app notice and the Runway Dev announcement are separate records.

  16. · Runway Dev adds Wan 3.0 generation

    The API generates up to 30 seconds at up to 1080p with generated audio, from text or a first frame. It also lists media references and endpoint frames. Runway uses wan3. This announcement does not list Alibaba's public webpage and document input.

  17. · Runway Dev catalog lists Wan 3.0 Prime

    The catalog lists wan3_prime for high-speed 2–30-second generation with audio and image, video, and audio references. August 28 is the catalog's Released field. This archive has no separate dated Runway Dev changelog entry for Prime.

Documentation and availability notes
  • Alibaba Cloud Model Studio · The guide covers adaptive endpoint frames, media references, editing, audio, and document or public URL inputs. Documentation examples can postdate the initial Model Studio release.
  • Runway Dev API · The catalog's Released values are August 26 for wan3 and August 28 for wan3_prime. The August 26 Runway Dev changelog names the standard model.
  • Alibaba Cloud Model Studio · The International release list dates Wan 2.5 Preview September 23, 2025. The mainland China section shows a September 19 entry. Each date belongs to its listed region.
Coverage through November 20, 2026. Exact publisher timestamps are shown in KST; date-only notices retain the publisher's day.Back to capability timeline