Changelog

2026-09-21

• Supplements the seedance-2.0-sg standard version 1080P / 4K two real-person reference images measurements: asset mode, 16:9, 4 seconds, and off-track, both of which were successfully decoded by downloading and full. The measured dimensions 1920×1080 and 3840×2160, the duration file is about 4.04 seconds; the record is not extended to other models or media assets combinations.
• Fixed missing resolution in SG 4K task queries. Successful tasks preserve resolution: "4k"; updated the 1080P / 4K examples, resolution labels and SG status definitions in OpenAPI.
• The seedance-2.0-sg image mix is repaired, and the request partial rejection of reference video was not provided, and the client can continue to use the original official compatibility parameters.
• 6 photo +1 audio examples to complement the officially content[]. This combination is enabled to produce real generation by 480p / 16:9 / 10 秒, generate audio, downloading and full decode; validation conclusions do not extend to other models or all media assets specifications.
• Supplements the difference between the request specification and the finished media information, with the example of synchronous OpenAPI. The existing definition of parameters and the reference media assets quantity ceiling remain unchanged.

2026-09-09

• Adds an example of a SG 4 model independent document, model matrix and OpenAPI, with 4 seconds, 2 diagram + 1 audio + 1 video reference generation; standard version supports 1080P / 4K, 2.5 supports 1080P.
• SG Single reference images role, output Token settlement, preauthorization balance return and direct access instructions for successful media URLs.
• Adds examples of seedance-2.0-nt, seedance-2.0-nt-fast, seedance-2.0-nt-mini, seedance-2.5-nt independent documents, navigation, model matrix and OpenAPI.
• Supplements 720P, examples of uniform parameters of type duration range, 4 seconds single-person net reference images and official compatibility parameters, and description of usage.total_tokens settlement and preauthorization differences.
• Clarified that 2.5 strict first/last-frame mode preserves the input aspect ratio, how to use successful result URLs, and that the public workflow for NT authenticated asset IDs is not yet verified.

2026-09-07

• Adds an example of an seedance-2.5-cl external model document, model matrix, model list and OpenAPI.
• Added the model's 4–30 second duration, 480P / 720P / 1080P resolutions, mixed references and authenticated Asset:// assets.
• Unequivocal authentication of media assets reference error returns the error directly and does not downgrade to a public network URL.

2026-09-05

• Supplementing the Seedance SZ official compatibility callback note: callback_url supports queued, running and final JSON callback, callback ID uses public task ID.
• Make it clear that the callback receiver should be judged by the status / error processing and result; the successful result URL may be a temporary signed media address and should be downloaded or re-loaded in a timely manner.
• The synchronous SZ model page, the Webhook Guide and the official path description maintain the consistency of content[], callback_url with the example of task query.
• Complements the unified multi-graph + multi-video + multi-sounding example for the original, KZ, CL, CL-B, GL and SZ series; the differences in models continue to be verified by the capacity configuration.

2026-09-04

• Add a new Seedance-2 CL-B open model line: seedance-2.0-cl-b-fast, seedance-2.0-cl-b-mini.
• The two additional models are used in /api/v3/contents/generations/tasks using the official content[]-shaped compatible description, and the parameters and capability boundaries are performed under the Volcengine-compatible Seedance 2.0 Fast/ Mini contract.
• role/alias protocol to supplement CL-B media entries, with model synchronous matrix, Feedance Guide and OpenAPI examples.

2026-09-03

• Documented Volcengine-compatible parameters for Seedance SZ: seedance-2.0-sz, seedance-2.0-sz-fast, seedance-2.0-sz-mini and seedance-2.5-sz accept content[], ratio, resolution, duration, generate_audio and related fields on /v1/videos, subject to each model's capability validation.
• request, query, list and cancel description for the additional compatible path /api/v3/contents/generations/tasks; the path uses the official model to name space, separate from the seedance-2.0-sz* public model name.
• Supplements the Asset:// certified photo, audio, video hybrid reference examples, and conflict rules content / references, duration / frames.

2026-09-02

• Add new public model documents for the Moon-2 AC series: moon-2.0-ac-* and moon-2.5-ac-*, to complement the resolution, duration, reference media assets boundary and by by sub/second billing degrees; the amount is based on price API and console.
• Adds a new Open Model for the Moon-2.5 GL series: moon-2.5-gl-a, using /v1/videos and references[].
• Support 480P / 720P, 4 to 30 seconds, six aspect ratio and up to 30 pictures, 10 videos, 10 audio, and a total of up to 50 reference media assets; audio reference needs to be accompanied by pictures or videos.
• The model is billing seconds at the actual price recorded in MoonNexAI console, prices API and billing.
• Clarified asynchronous image capabilities: requests use native asynchronous support verified for the selected model and group. If unavailable, the request fails before task creation or billing.
• The Gemini Photo asynchronous page continues to use the unified /v1/images/generations/async, task query and content API; size, aspect ratio, resolution and single generation numbers are based on current API Key and model capabilities.

2026-08-30

• Add a new Seedance SZ Public Series: seedance-2.0-sz, seedance-2.0-sz-fast, seedance-2.0-sz-mini, seedance-2.5-sz.
• Four models are used in a single /v1/videos and references[] format to support graphic, video and audio reference; single video reference capability has been validated.
• The SZ series is processed by billing on an official video-based basis when Zedance official Token uses billing, which contains video reference media assets; the actual price is based on the model list, console and billing records.

2026-08-22

• Add a new Moon-2 ME public series: moon-2.0-me-a-720p, moon-2.0-me-b-720p, moon-2.0-me-a-fast.
• The three models are used in a single /v1/videos and references[] format, supporting 720P, 5/10/15 seconds and five commonly used aspect ratio; A / A Fast supports up to 9 pictures, B supports up to 9 figures + 3 videos + 3 audio, and a total of up to 15 reference media assets.
• The ME series is billed per task and the actual price is based on MoonNexAI console, model list, price API and billing records.

2026-08-18

• Add a new Seedance-2 GL Public Series: seedance-2.0-gl, seedance-2.0-gl-fast, seedance-2.0-gl-mini.
• Three models are used in single /v1/videos and references[] to support 4 to 15 seconds, up to 9 pictures, 3 videos, 3 audio, and reference media assets totals up to 15; and back-up real-person reference, but content security clearance is still being performed.
• GL models accept stable public HTTPS URLs from your own storage. For reusable assets, /v1/assets/uploads can prepare Asset://asset_xxx references for the selected model. MoonNexAI-hosted URLs are an optional upload path for local files.
• Standard supports 480P, 720P and 1080P; Fast and Mini support 480P and 720P. All three are billed by output tokens. Check the MoonNexAI console and model-list API for current prices, groups and capabilities.

2026-07-30

• Add moon-2.0-ld-b: The capability is the same as moon-2.0-ld-a, supporting 720P, 5 to 15 seconds, up to 9 pictures, 3 videos, 3 audio, and a total of reference media assets up to 15.
• moon-2.0-ld-b is billing; the actual price is recorded in MoonNexAI console, prices API and billing.

2026-07-26

• moon-2.0-ld-a is available in the public model list, pricing API and documentation navigation. It supports 720P, 5–15 seconds, up to 9 images, 3 videos and 3 audio files, with at most 15 references in total.
• Video task queries now return directly to the security address that has been stored in the MoonNexAI media. media-nex.windfimusic.com is the official transferred media domain name managed by MoonNexAI; the client must preserve and use the complete URL containing the signature query parameters as it is.

2026-07-23

• MoonNexAI OSS test upload and media assets certification/preparation adjustment is free of charge, and the subsequent generation of task is still based on the corresponding model billing.
• Clarify that MoonNexAI OSS is for joint testing only, without a commitment to long-term file retention, long-term validity, continued availability or long-term service level; formal business and long-term use should be provided by the client for a stable media assets URL that is accessible on an anonymous public network.

2026-07-22

• Video references references[].alias are automatically created by type of media with 图片N, 视频N, 音频N symmetry when media assets is omitted.
• Built-in numbered aliases allow text immediately after the number, such as @图片1保持... or @图片 1保持...; the same applies to 视频N and 音频N. Custom aliases still require separation from following text.
• Supplements the Quest Note for invalid_reference_alias and duplicate_reference_alias; alias field is not @ and should remain the only one in the same request.

2026-07-21

• WC-F accepts up to 9 images, 3 videos and 3 audio files, with at most 12 entries in references[]. Audio must be accompanied by at least one image.
• WC-F continues to support text-only input, real people, 720P and 4–15 seconds. A 6 images + 3 videos + 3 audio files request passed asset validation and entered the queue, but did not produce a video. This does not establish reliable generation with the maximum reference combination.

2026-07-20

• Added generate_audio for seedance-2.0-cl, seedance-2.0-cl-fast and seedance-2.0-cl-mini. Audio is enabled when omitted; explicit false disables generated audio.
• OpenAPI synchronous generate_audio default and keep generateAudio compatible with field specifications.
• moon-2.0-wc-f-720p ' s reference media assets is adjusted to a maximum of 9 pictures and 3 audio, a total of up to 12 media assets; reference video is no longer supported.
• WC-F continues to support pure text, real content, 720P and 4 to 15 seconds; audio must be accompanied by at least one picture.

2026-07-19

• Add moon-2.0-wc-f-720p: Supports plain text, 720P, 4 to 15 seconds and real content.
• WC-F supports a maximum of 9 pictures and 3 audio, references[], a total of up to 12 media assets; without reference video, audio must be matched with pictures.
• WC-F is billing, the actual price is recorded in MoonNexAI console and billing.
• Adds the Moon-2 GL 720P series: moon-2.0-gl-a-720p and moon-2.0-gl-a-720p-fast, which supports pure text, character references, 4 to 15 seconds and 16:9, 9:16, 1:1 aspect ratio.
• The GL series supports a maximum of 4 pictures, 3 videos and 1 audio, with a maximum of 8 references to media assets; audio must be accompanied by pictures or videos, at a secondary price of console and at a price of API.
• Add moon-2.0-wc-e-720p: Supports plain text, 720P, 4 to 15-second target duration and real content.
• WC-E supports up to 4 pictures, 3 videos and 1 audio, for a total of up to 8 reference media assets; audio must be accompanied by pictures or video.
• WC-E is billed per task, the current price is based on MoonNexAI console, the price API and the actual billing record.

2026-07-17

• moon-2.0-wc-c-720p and moon-2.0-wc-d-720p are officially available and support pure text, 720P and 4 to 15 second target duration.
• WC-C supports up to 9 reference images, does not support video and audio; WC-D supports up to 4 pictures, 3 videos and 1 audio, which must be accompanied by a picture or video.
• Video task query and content downloads are uniformly returned to MoonNexAI ' s own controlled result address; only the full address returned by API is required for the client.

2026-07-16

• Fixing GPT Image asynchronous URL edit stability, single-graph URLs and multi-graph URL arrays can be submitted and queryed as asynchronous tasks processes.
• Supplement asynchronous Edit quality to choose field syntax: Use model default behaviour when omitting, with the visible value based on the selected model capability.
• Additional image content security failed processing statement: Read error.message when task fails, adjust the picture or prompt and create a new task.

2026-06-03

• synchronous MoonNexAI video gateway, customer assets and task chain documents:
  • Video guides complement the input boundaries of the customer public web URL, binary files, MoonNexAI hosting URL and model media assets ID.
  • The media assets guide supplements the /v1/assets/uploads behavior of mode=oss|asset|both and uploads the billing boundary prepared for media assets.
  • The asynchronous tasks guide supplements trace_id, request_id and task_chain.
  • Video model scenes supplement the media assets asset limits of some video models.
  • Updated OpenAPI examples and schemas for trace, task_chain, asset status and upload mode fields.
• Export to align with the original API document, the current OpenAPI has overwrited all 80 API operations in the export file.
• Adds a new API directory page, listing 84 MoonNexAI documents API operations by capacity.
• Add OpenAI-compatible, Photo and Midjourney, Web and Gemini-compatible Capability Guide.
• Adds access to the check list, asynchronous tasks and frequently asked questions page.
• Enhances the SDK example and the API Key security statement.
• Adds a new API directory overcheck to ensure that the API directory is consistent with OpenAPI.
• The API directory is automatically generated from OpenAPI to reduce subsequent maintenance deviations.
• Add an OpenAPI overcheck to prevent subsequent error and deletion of the API path.
• Extension of video generation, media assets resources, digital human, music / Suno competency guide to complement user call processes and examples.
• API Reference extends to 26 API paths such as media assets, digital human, Music task.
• Improved model, billing and SDK pages, with asynchronous task, asset upload and music examples.
• MoonNexAI API Docs first edition online.
• Adds fast start, authentication, model, billing, error code, Webhook, SDK example and API Reference.
• Add OpenAPI JSON entrance: /openapi/moonnexai.openapi.json.
• Add llms.txt and sitemap.xml, to facilitate access to public documents by search engines and AI assistants.