Home / By task / Make AI anime: five steps from character sheet to final cut

Free AI image generator8 Free AI cutout tools8 AI weekly reports6 AI text-to-video6 AI writing usable in China8 AI PDF summary6 Practise English speaking6 Make AI anime17 AI short-video scripts6 AI PPT8 AI product photos8 AI resumes6

Task guide · Creative production

Make AI anime: five steps from character sheet to final cut

Five steps and you have a file you can upload. It takes four things: a fixed character, a shot list, a renderer, and an edit that assembles and voices it. Order matters - generate before the character is fixed and every version will have a different face.

Explore the workflow Compare tools

17tools

14with a Chinese UI

5with an ICP filing

10with a free tier

01 / The process

A workflow you can actually follow

Treat AI anime as a chain of smaller jobs, not one tool that does everything.

STEP 01

Story and character

Write the biography, temperament and way of speaking as text first, then lock one reference image you compare every later shot against.

Deliverable: One fixed character description + one reference image

STEP 02

Storyboard

Break the story into 3-8 shots. Each row states shot size, camera movement, light, duration and line. Text only, no images here.

Deliverable: A 3-8 row shot list

STEP 03

Generate key frames

One key frame or one 3-5 second clip per shot. Budget three versions per shot - the first one is rarely usable.

Deliverable: One key frame or one short clip per shot

STEP 04

Animation and consistency

Keep the same character across shots. There are only three levers: a fixed model or style model, a fixed seed, a fixed reference image.

Deliverable: A consistent character + 3-5 second clips

STEP 05

Voice and edit

Two or three lines as one audio track, one caption track, shots cut in storyboard order, then one resolution, frame rate and codec on export. Keep each line under 15 Chinese characters so it can be timed to the picture.

Deliverable: One file you can upload

Make one 10-20 second test scene first: check whether the character drifts, whether motion is mush, and what it cost. Then scale up. Also - the tool database has no "export" stage, and inserting one would be padding; the usable export tools are the two GitHub projects in section 3 (MCP wrappers for ffmpeg and DaVinci Resolve).

02 / Tool finder

Choose tools for each stage

Start with the job you need to do. Check access, pricing and terms on the vendor page.

17 tools

WaWaXieZuo

Story & prompts

Write the character biography, temperament and way of speaking as text first; that text becomes the fixed part of your prompt.

Evidence: site HTTP 200 · 2 sources (ai-bot.cn,aigc.cn) · last checked 2026-10-16 · UI Chinese · price tier n/a · ICP filing: not recorded

wawawriter.comOfficial site ↗

Holopix AI

Story & prompts

Its stored one-liner says it is built for game, anime and illustration design - that is the use case.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI Chinese · price tier n/a · ICP filing: not recorded

holopix.cnOfficial site ↗

Chushou AI Painting

Story & prompts

Its stored one-liner describes a free AI drawing, model and sharing platform - you can reuse a style instead of describing one from scratch.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI Chinese · price tier n/a · ICP filing: not recorded

acgnai.artOfficial site ↗

Doubao AI

Story & prompts

Price field says free tier; fine for producing a structured shot table quickly.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI Chinese · price tier free tier · ICP filing: not recorded

doubao.comOfficial site ↗

DeepSeek

Story & prompts

Price field says free tier; stays on topic over long outlines.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI Chinese · price tier free tier · ICP filing 浙ICP备2023025841号

deepseek.comOfficial site ↗

JiMeng AI

Image & character

Stored one-liner: all-in-one AI video, image and digital-human creation - one entry point for stills and clips.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI Chinese · price tier n/a · ICP filing: not recorded

jimeng.jianying.comOfficial site ↗

KeLing AI

Image & character

Stored one-liner: Kuaishou's AI video generation tool. Price field says paid - read it before you top up.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI Chinese · price tier paid · ICP filing: not recorded

klingai.comOfficial site ↗

Vidu

Image & character

Price field says free tier and we recorded an ICP filing number, so the free allowance can test shots first.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI Chinese · price tier free tier · ICP filing 京ICP备2023012431号-9

vidu.cnOfficial site ↗

Manxiaoya

Image & character

Stored one-liner: a one-stop AI comic-drama platform - story to drama in one site.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI Chinese · price tier free tier · ICP filing 京ICP备2025129518号-2

manxiaoya.comOfficial site ↗

JuRiLu

Image & character

Stored one-liner: a one-stop AI anime video creation platform; price field says free tier.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI Chinese · price tier free tier · ICP filing: not recorded

ai.jurilu.comOfficial site ↗

LiblibAI

Animation

Stored one-liner: a one-stop AI content creation platform; price field says free tier - use it to lock one style.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI Chinese · price tier free tier · ICP filing: not recorded

liblib.artOfficial site ↗

Civitai

Animation

Stored one-liner: a free AI image and model sharing community; our UI check records an English interface.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI English · price tier free tier · ICP filing: not recorded

civitai.comOfficial site ↗

EbSynth

Animation

Stored one-liner: turns live-action video into painted animation - the exact move for carrying one key frame look across a clip.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI English · price tier free tier · ICP filing: not recorded

ebsynth.comOfficial site ↗

DomoAI

Animation

Stored one-liner: a one-stop AI video and animation platform - use it to unify a clip into one style.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI English · price tier free tier · ICP filing: not recorded

domoai.appOfficial site ↗

Qianyin Manyu

Voice & editing

Stored one-liner: a one-stop professional AI dubbing service, listing short video and audiobooks as its use cases.

Evidence: site HTTP 200 · 1 source (aigc.cn) · last checked 2026-10-16 · UI Chinese · price tier n/a · ICP filing: not recorded

qianyin123.comOfficial site ↗

XunFeiZhiZuo

Voice & editing

Stored one-liner mentions dubbing, speech synthesis and lip prediction; price field says free tier and we hold an ICP filing number for it.

Evidence: site HTTP 200 · 2 sources (ai-bot.cn,aigc.cn) · last checked 2026-10-16 · UI Chinese · price tier free tier · ICP filing 皖ICP备2023018870号-7

xfzhizuo.cnOfficial site ↗

Kuaijianji

Voice & editing

Stored one-liner: AI video editing with one-click assembly, auto captions and watermark removal.

Evidence: site HTTP 200 · 1 source (ai-bot.cn) · last checked 2026-10-16 · UI Chinese · price tier n/a · ICP filing 京ICP备2020049243号-6

kuaijianji.comOfficial site ↗

Editorial note: these entries come from our AI tool database, picked by hand per stage. They are not paid placements and not a ranking. Every value in the Evidence line is a stored field (site HTTP status, source count, last check date, UI verdict, price tier, ICP filing) - "site HTTP 200" only means we opened it that day, it does not mean we tested it. The online/local column is our own classification (cloud service = online, download and install = local, useful both ways = shown on both routes); it is not a stored field. Pricing, allowances and regional access: check the vendor page.

03 / Resource library

The files and docs behind the workflow

Animation consistency, models, nodes and export live as GitHub projects and MCP servers, not as product homepages.

The AI tool database has no "local workflow" category - these things exist as GitHub projects, MCP servers and agent skills, so they live in our other database. The eight entries below are real: the star count is public, the date is when we last saw it, and every link goes to the project page.

awesomeComfyUI

Build and run Stable Diffusion workflows using a node-based visual interfaceMCP servers

yesterday

awesomeConstantineB6/comfy-pilot

Lets you view, edit and run ComfyUI node-based image generation workflows, with an embedded terminalMCP servers

232 ★

yesterday

awesomeAIDC-AI/Pixelle-MCP

Omnimodal AIGC framework converts ComfyUI workflows into MCP tools for text, image, and other mediaMCP servers

1,129 ★

yesterday

skillsmpcomfy-mcp-fixes

Resolve issues running a local ComfyUI instance via comfy-mcp, comfy-cli, and an in-app browser

yesterday

skillsmpcomfy-duplicate-model-roots

Fix duplicate, empty, or conflicting model-root config problems in ComfyUI launchers and package managers

yesterday

skillsmpai-image-prompts-skill

Get curated prompts from a 10,000+ real-world image generation prompt library that works with any AI image tool

1,248 ★

yesterday

awesomevideo-creator/ffmpeg-mcp

Local video search, trimming, stitching and playback through conversation lets you edit and play videos by typing commandsMCP servers

153 ★

yesterday

awesomesamuelgursky/davinci-resolve-mcp

Integrate DaVinci Resolve for video editing, color grading, media management and project controlMCP servers

3,662 ★

yesterday

How we verify a resource: a source link, when we last saw it, whether a README exists, and the public star count - all four are on the card for you to check. The database has no version, licence or compatibility columns, so this section does not print them; printing them would be inventing them.

04 / Prompt starter

Start with a reusable prompt

A structure, not a magic spell. Replace the bracketed parts and test it in your own model.

Every block below is an example prompt we wrote. None of them is an official prompt from any project, and none was copied. Use it as a structure: replace the bracketed parts.

1. Character reference sheet (locks the character)Example we wrote

Create a clean anime character reference sheet on a white background, front view.
Character: [age/gender], [hair style and colour], [eye colour], [outfit and palette], [one distinctive accessory].
Requirements: full body, standing straight, hands relaxed, calm expression; clean line art, cel shading; this character only.
Negative: extra fingers, text, watermark, a second character, busy background.

Why it is written this way: it separates the character from the background. The sheet only carries face and outfit; describe the scene later, in the generation step.

2. Shot list (text only, no images)Example we wrote

You are an animation storyboard artist. Break the story below into 3 shots and output a table with these fixed columns: shot number / shot size / what happens / camera movement / light / duration (seconds) / dialogue.
Rules: one event per shot; total duration under 20 seconds; each dialogue line under 15 characters.
Story: [one sentence, e.g. a girl finds a glowing cat in the rain].

Why it is written this way: fixing the column names is what makes the output paste straight into the next step, and "one event per shot" stops the model splitting one action in two.

3. Consistency check (compare two versions yourself)Example we wrote

Below are two versions of the same character. Compare them item by item and point out the inconsistencies. Answer only "consistent / not consistent + exactly what differs":
1) face shape and proportions 2) hair length and colour 3) outfit cut and palette 4) accessory placement 5) overall style (line weight, shading method).
Version A: [paste the first description or prompt]
Version B: [paste the second]

Why it is written this way: a model will not look at your images, but it will compare two descriptions. Five fixed criteria beat asking "do these look the same".

05 / First session

A low-risk way to get started

Prove the pipeline on one small scene before you spend time or credits on a full episode.

The goal is not an episode. It is to prove the chain works on this machine and this account. All four steps use one character and one location.

Write a 3-shot micro-story

One character, one location, one action. Use the storyboard tools and prompt 2 from section 4 to produce the table.

Deliverable: A 3-row shot list

Produce one reference sheet

Use the story and character tools and prompt 1; render 3 versions, keep the closest one, and save its prompt.

Deliverable: One reference image + one fixed character paragraph

Render one frame per shot

Use the reference image, run the same prompt on all 3 shots. Do not aim for perfect yet - only check whether the character changed.

Deliverable: 3 frames (or 3 three-second clips)

Voice it, cut it, export it

Dub two lines, cut in storyboard order, normalise resolution and frame rate, then export.

Deliverable: One 15-30 second test clip

06 / Common questions

Before you begin

Do I need a powerful computer?

Not for hosted tools. Local generation depends on the model, resolution, workflow and available memory - ComfyUI's own README states a floor of 4GB VRAM + 8GB RAM (their words, not a recommended configuration), and the tighter the memory the slower the render.

Can one tool make a whole episode?

Some products combine several steps (the generation stage below lists all-in-one platforms), but long-form work usually separates story, character, shots, animation and editing. Test continuity before scaling up.

Is everything free for commercial use?

No. Model weights, style models (LoRA), assets and hosted services carry separate terms: being able to render with a model does not mean its licence allows commercial use, and a hosted service adds its own terms. Open each source page before you publish or sell - we flag it, we do not decide it for you.

Can I use these tools in China?

Access and payment differ by service and region and they change. This page records two things we checked - the UI language verdict and the ICP filing number - and it has <b>no</b> "accessible in China" column, because the database has no such field. There is no "works in China" badge on any card; a badge for data we do not have would be invented.

Why is there no score on the cards?

Because the database has no such field. What we store is: whether the site answers (HTTP status), how many sources mentioned it, when we last checked it, whether the interface is Chinese, whether we found pricing evidence, and the ICP filing number. A score without data behind it is worse than none.

Why does the price say n/a?

Because we found no clear pricing evidence (neither the homepage nor a /pricing page). That does not mean paid and does not mean free - the vendor page is the source of truth.

Are these official prompts from some project?

No. Every block in section 4 is labelled as an example prompt we wrote. Its job is to show how to describe a character and a shot, not to quote any project.

Where does this usually break?

Two places. (1) Putting face, hair, outfit, accessories and background into one prompt means the image satisfies two or three of them - fix the character first and add the scene at the generation stage. (2) Rewriting the description in each new tool changes the character immediately - keep it as one fixed paragraph and paste it unchanged. Both are our own summary, not an official statement from any vendor.

What we actually check

Every label here - Chinese UI, ICP filing, free tier - comes from a field we recorded per tool. Where we found no evidence we write n/a rather than infer one. Pricing should be checked on the vendor page: what we show is what we saw on the day we looked.

Found something missing or wrong?

Tell us and we re-check it against the source before it changes.

Have a workflow or resource to add?

Include a source link, the version and the environment - we only take what can be checked.

Suggest a resource ↗

← Practise English speaking · task 8 of 12 · AI short-video scripts → · 8 catalogue entries