Xiaoyunque AI

What is Xiaoyunque AI?

Xiaoyunque AI (小云雀AI) is a Chinese-language AI creation platform, available on the web and in iOS and Android apps, that turns prompts, scripts, images and reference videos into short videos, multi-episode short dramas, images and advertising creatives. Its homepage calls it Douyin's content-creation Agent, and the site footer describes it as ByteDance's creative Agent platform.

The user agreement names Shenzhen Lianmeng Technology Co., Ltd. (深圳市脸萌科技有限公司) and its affiliates as the owner and operator of the Xiaoyunque app and official website. The service runs on a jianying.com subdomain, and the vendor's algorithm filing notice says a Jianying effects algorithm is mainly applied to Xiaoyunque's AI effects features. Beyond that, the official pages reviewed do not define how Xiaoyunque relates to the Jianying product.

The product centres on story-driven video, with a short-drama agent, a marketing agent and general agent tools running on Seedance video and Seedream image models. Generation is paid for in credits under yuan-priced plans, and the service is set up for mainland China, with real-name phone verification and data stored in China.

Core features

Short-drama and comic-drama agent

The short-drama agent is an AI short drama generator that works at the level of a whole series rather than single clips.

• Script-to-episode generation: the agent is described as parsing a script automatically and generating it continuously, so a script alone can be turned into complete episodes, while characters and scenes for the entire series are produced in one pass with each character's styling mapped across different times and settings.
• Script Assistant: conversational co-writing supports writing from scratch, novel adaptation, serialized series, rewriting into narrated drama and localization for overseas audiences.
• 3D director stage: users place character and prop models in a 3D space, adjust their positions and poses, set camera positions and send screenshots back to the canvas as composition references for storyboards or video generation.
• Smart preview: the tool previews a whole group of shots, plans the storyboard as a set and inserts key frames into the matching segments, which the vendor says sharply reduces rerolls.

A story-specific infinite canvas serves as a visual workspace for organizing and generating characters, scenes and related reference material, so assets made in one step can be reused in later ones.

Marketing agent

• Brief-to-asset generation: from a product and its selling points, the marketing agent finds a direction through Showcase examples and creative, hook and style libraries, then generates multiple creative scripts and assets.
• Marketing skills: the agent can call batch generation, viral-asset variation, video translation and business-growth analysis skills.

For sellers, this makes Xiaoyunque AI an AI video ad maker covering the ad concept, opening hook and visual style, with results collected on a canvas for further editing.

General creation modes

• Agent mode: a single request can target a marketing poster, a 15-second short or a one-minute brand story film, with the agent interpreting the intent, planning the creative path and choosing models.
• Immersive short films: this mode supports text-to-video, image-to-video, video-link references and mixed image-plus-video references.

Models

• Seedance 2.5: each run generates 4 to 30 seconds and videos can be extended, while one job can combine up to 30 images, 10 videos and 10 audio clips, with at most 50 reference assets in total.
• Resolution: 1080p output went live on August 14, and a route that generates at 480p and then upscales supports super-resolution up to 4K.
• Model catalogue: the model page lists Seedance 2.5, 2.0, 2.0 fast and 2.0 mini for video, Seedream among 10+ image models, and Seed Audio models covering 20 languages.
• Disclosed model filings: the vendor's filing notice lists the Doubao large language, text-to-image and video generation models for AI content generation, plus DeepSeek Chat for AI writing and conversational creation.

Guide

Making a short drama with the agent

1. Upload a script: open the short-drama agent and upload a prepared script in txt or docx format, or paste the text, with scripts of up to 100,000 characters supported; alternatively, describe a story idea and let AI write the script.
2. Confirm the outline and settings: review and edit the generated script summary, then fill in the global settings form with a project name, visual style and aspect ratio.
3. Build the character library: the agent extracts every character and their styling over time, and editing a character's image or voice in the asset library updates every episode in which that character appears.
4. Generate scenes: all scenes used across the series are generated together and can be regenerated by editing their descriptions in the library.
5. Split shots and generate: each episode is broken into shots with storyboard scripts; mention library characters and scenes with @ in the storyboard, then batch-generate several episodes and shots at once.
6. Composite and export: merge the generated clips and export the full series.

If a character drifts between shots, the manual's FAQ recommends @-mentioning the character image inside that specific shot's storyboard to keep the look consistent.

Use cases and examples

• Overseas short-drama remakes: the remake and redraw upgrade adapts dramas for overseas markets with episodes of up to 5 minutes and 100 minutes in total, backed by white-model references, a library of more than 7,000 characters and localized overseas scripts.
• Cross-border ad localization: the video translation skill is described as translating an original video into multiple languages and faces while keeping its selling points and pacing, and the manual's examples include UK and Thailand localized versions.
• Product ads from a short brief: the manual's sample brief is a portable juicer with three selling points, aimed at commuters and gym-goers, for a 15-second Douyin placement with a light, authentic product-seeding tone.
• One-take promotional clips: the one-take tool accepts 2 to 10 images, recognizes their content, fills in the transitions and generates a continuous long shot.

Who is it for

• Creators and content teams: the web manual addresses individuals such as self-media accounts, short-video bloggers and people recording daily life, as well as professional teams including short-drama creators, content studios and brand marketing teams.
• Beginners and prepared creators alike: Agent mode is pitched at people with no storyboard skills who start from one sentence, people who already have a script, people bringing their own images, videos or copy, and people chasing a specific style.
• Adults, or minors with consent: the user agreement says the main audience is adults, and users under 18 should read it with a guardian and use the service only with guardian consent.

It is a weaker fit for buyers who need billing outside yuan-priced plans or who cannot complete the real-name verification described under Limitations.

Platforms

• Web: the manual positions the web version for large-screen control, running several creation tasks in parallel and batch production for teams and creators who publish continuously.
• Mobile apps: the official site offers iOS and Android downloads alongside the web workspace.
• iOS listing: the China App Store lists the app for iPhone only, free to download with in-app purchases and rated 13+.
• App rating: as captured on October 4, 2026, the China App Store listing showed 4.9 out of 5 from about 64,000 ratings, an aggregate score rather than a verified review sample.
• Sign-in: accounts can be created through a third-party account such as Douyin, which shares profile details and an authorized phone number, and the phone number may need to be re-verified.

Pricing

The pricing page shows annual plans by default, all in Chinese yuan:

• Renewal: every paid plan renews automatically, and the pricing page states it can be cancelled at any time.
• Free tier: free users claim daily free credits by logging in, which the page presents as the first step into video and image creation.
• Trial model: the model menu offers non-members a limited number of Seedance 2.0 Mini trial runs, at a promotional rate as low as 4 credits per second.
• Paid benefits: on every paid plan, credits gifted each day expire the same day, extra credits can be topped up, the brand watermark is removed from finished works, and batch creation allows more simultaneous generation tasks.
• Billing cycles: quarterly and monthly auto-renewal tabs sit next to the default annual view summarized above.
• App Store purchases: in the China App Store, in-app purchases include a monthly Basic membership at ¥79.00 and credit top-ups starting at 500 credits for ¥50.00.
• Current promotion: a promotion running from September 22 to October 8 discounts the plans and lists Seedance 2.5 720p generation from ¥0.28 per second and Seedance 2.0 VIP from ¥0.44 per second.

Premium also offers a credit slider with tiers from 6.3k to 27.7k credits a month; the table shows the default tier.

Xiaoyunque AI alternatives

Media coverage published in June 2026 noted that the video generation in Doubao, Jimeng, Jianying, Suibian and Xiaoyunque all ran on the same Seedance 2.0, with differences limited to available functions, generation speed and quota.

• WonderClip (Alibaba Cloud): an AIGC video platform built around three core modes, storyboard, infinite canvas and Agent, covering industrial-scale short-drama production through marketing creatives.
• WonderClip entry offer: new users receive 200 credits plus five days of Standard membership, and paid membership adds 4K image and video generation, faster generation, team collaboration and API calls.

Both products center on storyboard-driven short-drama production; the WonderClip page reviewed stresses team collaboration and API access, while Xiaoyunque AI's manual stresses the 3D director stage, overseas remakes and Douyin-oriented ad briefs.

Limitations

Access and account requirements

• Real-name verification: users can generate, create and publish audio, video, images and text only after completing real-name identity verification as required.
• Phone number: the privacy policy calls the phone number necessary for real-name rules and says refusing it may mean no AI content generation, publishing or interaction features.
• Jurisdiction: unless otherwise agreed, the operator does not guarantee that the service is lawful or appropriate outside the jurisdictions where it is officially released, and users elsewhere bear the risk of complying with local law.
• Governing law: the agreement is governed by the law of the People's Republic of China, with disputes going to a court in Haidian District, Beijing.

Output and quality limits

• AI labels: the operator may add AI-generated content labels under Chinese labeling rules and standards, and users must not cover, alter or remove them.
• No accuracy guarantee: the operator says it cannot guarantee that outputs are true, accurate, reliable or unique, and the same or similar inputs may produce the same or similar results.
• Long clips need explicit pacing: the Seedance 2.5 manual warns that 30-second clips carry more information, so prompts must spell out the pacing or the second half may drift off point.
• Reported friction: June 2026 media coverage, citing feedback from many users while the service ran Seedance 2.0, reported no fine shot-level control, limits on duration and resolution, style inconsistency across segments, weak support for complex narratives and peak-time queues of up to eight hours.

Rights and privacy

• Ownership: to the extent laws allow, intellectual property and other property rights in inputs, generated content and published content belong to the user or the lawful rights holder.
• Service and model license: users grant a worldwide, free, non-exclusive, sublicensable license to use inputs, outputs and published content to provide the service, optimize products and models, and keep the service secure; users who do not want this can contact the operator through the feedback page or by email.
• Published content: for content published through the service, users also grant a worldwide, free, non-exclusive, sublicensable license that covers promotion, advertising, marketing and research uses.
• Remixing by others: uploaded and published content, including voices, audio and dialogue, may be used by other users through features such as "make the same", reference images and other AI generation, and the results can be published, reposted, copied and downloaded.
• Commercial avatars: the App Store description markets the digital-human avatars as commercially usable; in the agreement reviewed, no separate clause on commercial licensing of generated videos was found.
• Model optimization: the privacy policy says collected data and the corresponding model outputs may be used for model optimization after encryption and strict de-identification that prevents re-identification.
• Data location: personal information collected and generated during operations inside the People's Republic of China is stored inside China.

FAQ

Q1. Is Xiaoyunque AI free to use?

Partly. Logging in grants daily free credits and non-members get limited Seedance 2.0 Mini trials, while paid web plans start at ¥759 per year for Basic and the China App Store sells a monthly Basic membership at ¥79.00.

Q2. Can I use Xiaoyunque AI outside mainland China?

The official pages reviewed do not say whether the service is offered outside mainland China. The terms disclaim lawfulness outside officially released jurisdictions, accounts require real-name phone verification, and the iOS app found during research is listed in the China App Store.

Q3. Does Xiaoyunque AI use my content to improve its models?

The user agreement's license covers optimizing products and models, and the privacy policy allows de-identified data and outputs to be used for model optimization. Users who object can contact the operator through the feedback page or by email.

Q4. Who owns the videos I generate?

To the extent laws allow, rights in inputs and generated content belong to the user or the lawful rights holder, but published content is also licensed to the operator for promotion, and other users may remix it through in-app features.