小云雀

1.05
A content-creation Agent for short-form videos and AI dramas that can automatically handle scripts, characters, storyboards, video generation, and basic editing from a theme or existing screenplay.
Advertisement 728 × 90
Company字节跳动
CategoryAI Video
Released2025-06
Updated2026-09-01

小云雀 Overview

When Xiaoyunque first launched, its main selling point was “generating a video from one sentence.”

By 2026, its focus had increasingly shifted toward short-drama production.

The short-drama Agent launched in March 2026 integrates Seedance 2.0. It can read a script directly, identify the characters, divide the story into shots, and generate the video section by section.

Integration with Seedance 2.5 began at the end of July, extending the maximum length of a single video generation to approximately 30 seconds.

The underlying models mainly come from ByteDance’s Doubao and Seedance families, while DeepSeek is also used for semantic understanding.

In practice, users do not necessarily need to begin with a prompt.

They can provide a theme and ask the system to write the script, or upload an existing script and let the Agent continue producing content according to the story.

Reference images and videos can also be provided so the model can follow the character’s appearance, visual style, and camera movement.

The difference between Xiaoyunque and an ordinary text-to-video tool is therefore not simply the addition of more model-selection buttons.

It is closer to a platform that wraps a complete content-production workflow around the underlying models.

小云雀 Pricing

PlanPriceDescription
Free Plan Free The free Credit allowance may change over time. During the platform’s early launch period, it offered approximately 120 Credits per day for testing basic generation features.
Membership Plan Around $11/month Includes approximately 830 Credits per month, with first-month discounts available during certain promotional periods.
Credit Packages Pay-as-you-go Additional Credits can be purchased separately when the allowance included with the membership runs out.

Xiaoyunque uses a membership-and-Credits pricing model, with video generation accounting for most of the usage.

Credit consumption varies significantly between models.

Seedance 2.0’s Standard mode reportedly consumes approximately 14 Credits per second, while Fast mode uses around 11 Credits per second.

A 30-second 720p video generated with Seedance 2.5 reportedly costs approximately 780 Credits, equivalent to around 26 Credits per second.

At that rate, 830 Credits do not go as far as they may initially appear.

This is especially true when producing short dramas, because a shot is rarely satisfactory on the first attempt. If the character drifts, the movement is incorrect, or the composition is off, regenerating the shot consumes another full round of Credits.

The membership fee is therefore only the entry cost. The number that truly needs to be calculated is the number of regenerations.

Public cost estimates place the tool expense for one minute of video at several dozen yuan. Once production expands to dozens of episodes, the total can quickly reach several thousand yuan.

小云雀 Key Features

  1. Smart video creation: Enter a topic, written requirements, or reference materials to automatically generate a script, storyboard, and video assets, then combine them into a short video.
  2. Short-drama Agent: Upload a .docx script to automatically analyze the plot, characters, and settings, divide it into storyboard shots, and generate the episode section by section.
  3. Seedance video generation: Uses models such as Seedance 2.0 and 2.5 to produce shots, with support for longer video generation and continuous content creation.
  4. Character and style references: Add character images, visual references, and video materials so subsequent shots can maintain the same characters and visual direction as closely as possible.
  5. AI avatar videos: Select a digital presenter, enter a script, and quickly generate a talking-head video.
  6. AI image and design tools: Supports common image-generation and editing tasks such as text-to-image creation, poster generation, and automatic background replacement.
  7. Reference-video recreation: Analyzes the shots, pacing, and visual language of a reference video, then generates new content with a similar structure.

小云雀 Editorial Review

The most convenient thing about Xiaoyunque is that you do not need to divide the entire video-production process across a dozen separate tools.

I tested it with a wuxia short-drama script.

After the script is uploaded, it first analyzes the characters, settings, and story before breaking the content down into individual shots.

This approach is much more reliable than sending the entire block of text directly to a video model.

One of the most common problems with video models is that every shot can look as though it was produced independently.

The protagonist may wear black in one shot and suddenly change clothes in the next. A character who had long hair may have a different face as soon as the camera angle changes.

Xiaoyunque organizes the character information in advance and passes those settings into subsequent shots.

Character consistency is noticeably better than when every shot is generated completely independently.

However, this does not mean that locking a character guarantees there will never be any visual drift.

Faces, clothing, and other details still need to be checked in complex scenes, during large character movements, or when a shot is regenerated repeatedly.

Reference materials are often more useful than prompts alone.

If you want a particular style of camera movement, you can provide a reference video.

If you want to maintain a consistent visual style, you can upload several reference images at once.

This approach aligns well with how ordinary creators work.

It can be difficult to describe exactly what kind of cinematic language you want, but a visual reference can communicate it immediately.

The factor most likely to discourage users is still Credit consumption.

Once you begin generating video repeatedly, the balance decreases very quickly.

In one publicly shared test, a user uploaded a script of approximately 6,000 Chinese characters and purchased 1,200 Credits. The entire allowance had been consumed by the sixth storyboard shot of the first episode, representing around 47 seconds of content.

This also shows why short-drama costs cannot be calculated solely from the length of the final video.

A finished 60-second video may involve two or three minutes—or even more—of rejected footage generated behind the scenes.

Queueing is another issue.

Waiting during peak periods is not unusual after a popular model becomes available. If a team already follows a regular publishing schedule, generation speed itself must be included in the operational cost.

Xiaoyunque is therefore not a tool where you click once and wait for a complete short drama to appear.

The script, characters, and storyboards still require human review during the early stages.

If the direction is wrong at any of those points, the Agent will continue generating along the wrong path while consuming Credits at the same time.

Pros

  • A fairly complete script-to-video workflow: There is no need to move repeatedly between separate script, storyboard, and video tools.
  • Strong focus on short dramas: Characters, storylines, and consecutive shots are easier to manage than with standalone text-to-video generation.
  • Useful reference materials: Images and videos can communicate a visual style directly instead of relying entirely on prompts.
  • Fast Seedance integration: New ByteDance video models generally enter the product workflow relatively quickly.
  • Accessible to beginners: Even users without video-production experience can follow the Agent through a complete workflow.
  • Suitable for rapid proof-of-concept videos: Teams can produce an initial version at relatively low cost to determine whether a story is worth developing further.

Cons

  • Video consumes Credits rapidly: A few repeated generations can quickly exhaust the monthly allowance.
  • Character consistency is not completely stable: Complex shots can still produce changes in characters or clothing.
  • Queues may occur during peak periods: This can disrupt high-frequency production schedules.
  • Fine control remains limited: Localized movement, complex fight scenes, and precise shot adjustments are less controllable than in professional production software.
  • AI can misinterpret the intention: Complex plots and emotional scenes may produce shots that do not match the original meaning.
  • Costs must be calculated before scaling production: A single finished minute may require a substantial amount of unusable generated footage.

Best for / Not ideal for

Best for

  • AI short-drama creators: Suitable for people who already have scripts and want to produce characters, storyboards, and an initial video quickly.
  • Short-form video teams: Useful for continuously testing different genres and creative directions.
  • Marketing and e-commerce operators: Suitable for frequent production of presenter videos, product demonstrations, and advertisements.
  • People without professional video experience: Allows users to take an idea through the entire workflow.
  • Teams that need rapid proof-of-concept videos: Useful for validating the story and visual direction before committing to full production.

Not ideal for

  • People with extremely limited budgets: Frequent generation and repeated attempts can be expensive.
  • Film teams requiring complete control over every shot: Professional projects still require human direction, editing, and post-production.
  • People unwilling to review intermediate results: Errors in characters or storyboards can continue affecting every subsequent stage.
  • People who require consistently stable output: AI video generation still involves significant randomness.
  • People who only occasionally generate a single shot: The complete Agent workflow may feel unnecessarily heavy.

Summary

Xiaoyunque is not simply trying to provide a web interface for Seedance.

It is attempting to connect scripts, characters, storyboards, video generation, and editing into a single short-drama production pipeline.

For users who previously had to move assets back and forth between several AI tools, this integration is genuinely convenient.

It is particularly well suited to short dramas.

You can provide a script and let the system identify the characters and divide the story into shots before generating them step by step. This is considerably easier than starting every shot from a blank prompt.

However, it has not yet reached the point where it can produce an entire short drama with one click.

Characters may drift in appearance, complex movements may fail, and some shots may need to be regenerated. Every regeneration consumes additional Credits.

For your first project, do not immediately submit a script containing dozens of episodes.

Start with a story segment lasting less than one minute.

Complete the full workflow—from characters and storyboards to the final video—then check how many shots need to be regenerated and how many Credits the process actually consumes.

If the cost and number of revisions for a one-minute sample remain acceptable, then consider a membership and larger-scale production.

For Xiaoyunque, the question is no longer whether it can generate video.

The more important question is: Can it consistently produce enough usable shots within a budget you are willing to pay?

Comments (0)

Leave a comment

Advertisement 728 × 90

Similar Tools

Gemini Omni
88
A multimodal AI model that can understand text, images, audio, and video, then generate and refine videos through natural-language instructions.
AI Video
Rive
88
A production-grade animation tool that brings design, animation, and interaction logic into a single file, ready to run directly in apps, websites, and games.
AI Video
Face Swap by Akool
87
An AI face-swap tool focused on realism and multi-person swaps. You can replace faces simply by uploading photos, but video quality and commercial licensing terms require extra attention.
AI Video
Runway Gen-4.5
84
An AI video model focused on realistic motion, camera control, and character consistency. It’s starting to move beyond simply generating good-looking clips and toward becoming a more serious creative tool.
AI VideoAI comic drama
Neural Frames
81
An AI video tool that makes visuals move with the music. Rather than simply generating video, it focuses on syncing motion and visuals to the beat and soundtrack.
AI Video
即梦AI
80
An AI creative tool that understands natural language, researches information, and turns ideas directly into images or videos.
AI VideoAI Image
Synthesia
77
An AI video platform for corporate training and internal communications that quickly turns text, documents, or webpages into digital-avatar presentations, with support for multiple languages, collaboration, and LMS workflows.
AI Video
Descript
76
Edit videos like you edit a document—video editing no longer has to be a technical skill.
AI Video