Why Everyone’s Suddenly Watching AI-Made 2D Story Videos (And How You Can Make One Tonight)
Scroll through Reels or Shorts for more than five minutes and you’ll run into one of them — a short, hand-drawn-looking cartoon about a farmer who got wronged, a son clearing his father’s name, or a poor widow finally getting justice. No big studio, no Pixar budget. Just a phone, a couple of free AI tools, and a story that hits close to home.
I got curious about this trend the same way most people do — I watched three of them back-to-back at 1 a.m. and thought, “wait, how is one person making all of these?” Turns out, the answer isn’t a secret animation team. It’s a workflow. And once you see how the pieces fit together, it’s genuinely something you can build from your phone this week.
This guide walks through that exact workflow — from picking a story to uploading a finished video — and explains why each step exists, not just which buttons to tap.
Why This Format Is Working Right Now
Before we touch any tools, it’s worth understanding why these videos are pulling in millions of views, because that understanding is what will actually make your videos good — not just technically correct.
The stories are emotionally simple, and that’s the point. Justice, betrayal, family loyalty, a wrong finally set right — these themes don’t need translation. A 90-second story about someone’s goats being stolen and a corrupt village head covering for the real thief works in Karachi, in Delhi, and in a diaspora living room in London. You don’t need clever writing. You need a clear villain and a satisfying ending.
The 2D style is a feature, not a limitation. Hyper-realistic AI video still trips over the “uncanny valley” — faces that are almost right but not quite, which makes viewers uneasy. Stylized 2D sidesteps that completely. Audiences read it as an artistic choice, not an AI shortcut.
Language does a lot of the work. Most of the biggest channels in this space narrate in Roman Urdu or Hindi rather than English. That’s not a limitation — it’s the whole reason these videos outperform English-language competitors for South Asian audiences who’d rather hear a story in their own voice.
Each video is a complete, bite-sized story. No cliffhangers, no “part 2 coming soon.” Beginning, middle, end, in under two minutes. That completeness is exactly what keeps people scrolling to the next one instead of bouncing — and that retention pattern is what the algorithm rewards.
None of this replaces a real animation studio, and it’s not trying to. It’s built for volume and consistency on social platforms, where a decent story told often beats a perfect story told rarely.
The Toolkit: What You Actually Need
You don’t need to learn four apps at once. Here’s what each one is doing in the pipeline, and why it’s there.
A language model — ChatGPT or Claude. This is where your story actually gets built: the plot, the characters, and the scene-by-scene breakdown that everything else depends on. If your prompt here is weak, nothing downstream can save the video. This is the step people skip and regret.
An AI video generator — Google Flow (or a similar Kling/Flow-style tool). This does the visual heavy lifting: character reference sheets, background art, and the actual 10-second animated clips. The feature that matters most here is the one that lets you “attach” a saved character to a new scene, so your protagonist doesn’t randomly change faces halfway through the story.
A phone video editor — CapCut, InShot, or whatever you’re comfortable with. AI gives you raw clips; editing gives them a heartbeat. Pacing, music, and voiceover are what turn six unconnected clips into something people actually feel.
A voice recorder or text-to-speech app — for narration. Most channels in this niche don’t bother with AI lip-sync. A narrated voiceover over 2D visuals is the standard, and honestly, it’s more forgiving to produce.
Building the Video, Step by Step
1. Get Your Story Idea Right Before Touching Any Tool
Open ChatGPT or Claude and ask it to generate several story concepts for your niche — moral lessons, village drama, family conflict, whatever direction you want. Ask specifically for Roman Urdu or Hindi output if that’s your audience, and ask for a clear protagonist, a conflict, and a resolution for each idea.
Here’s the part beginners get wrong: they take the first idea the model gives them. Don’t. Read through all of them and pick whichever one actually makes you feel something. If a story doesn’t move you while reading a two-line summary, it won’t move a stranger scrolling past it in half a second.
If you’d rather write your own concept, you barely need more than three or four sentences — something like: a farmer’s two goats go missing, the village blames an innocent laborer, and the farmer’s son stays up at night to find out the village chief’s own servant did it. That’s plenty for the AI to expand into a full scene breakdown.
2. Decide Your Length in Scenes, Not Minutes
This workflow has one fixed rule: each scene becomes a 10-second clip. So instead of thinking “I want a 90-second video,” think “I need 9 scenes.”
For your first few videos, stay in the 8–10 scene range (roughly 80–100 seconds). It’s tempting to go big and plan a 20-scene, 3-and-a-half-minute epic right out of the gate — I’d steer clear of that until you’ve got a couple of finished videos under your belt. Longer videos are harder to pace, cost more generation credits, and are more likely to end up half-finished in a folder somewhere.
3. Get Your Character and Background Descriptions
Once the scene count is set, your AI will hand you two lists: character descriptions (appearance, clothing, age, expression) and background descriptions (the farmer’s home, the goat pen, the village meeting spot, a dark crossroad).
Don’t rush this part. If your character description feels generic, push back and ask for more specific detail — age, exact clothing, a defining expression. An hour spent nailing down what your protagonist looks like will save you from the much bigger headache of him looking like a different person in every third scene.
4. Build Character Sheets
This is where text becomes image. In Google Flow, head to the Characters section, pick a stylized 2D model (Nano Banana 2 works well for this look), and generate each character one at a time from their description. Once you’re happy with how someone looks, save and clearly label them — “Rahmat_Farmer” beats “Untitled_4.”
Think of this step as casting. Once “Rahmat” exists as a saved character, you can drop him into any scene wearing the same clothes with the same face — which is really the entire difference between a video that looks intentional and one that looks like it was assembled by six different artists who never spoke to each other.
One thing worth locking in early: use the same model and settings for every character in a story. Mixing models mid-project is the fastest way to end up with an art style that visibly clashes from one character to the next.
5. Create Your Backgrounds
Switch to the image or scenes section, keep the same model selected, set the aspect ratio to 16:9, and generate each location from your background descriptions.
Backgrounds do more work than people expect — they tell the viewer’s brain “we’re back in a place we recognize,” which makes the story easier to follow without a single word of explanation. A small trick that helps a lot: add a mood detail to the prompt, like “dusty evening light” or “monsoon clouds gathering.” That one phrase is often the difference between a flat background and one that actually feels like part of the story.
6. Turn Your Story Into Scene-by-Scene Prompts
Back in your language model, ask it to break the story into individual scene prompts. A good scene prompt doesn’t say “a man is sad” — it says something closer to “Rahmat sits on a wooden cot, head in his hands, shoulders shaking slightly, dim lantern light.” That level of detail is what gives the AI video model something to actually animate.
Each scene prompt should also tell you which characters and which background belong in it — pay attention to that list, because it’s your checklist for the next step.
7. Generate the Actual Video Clips
This is the main event. In Google Flow’s video section, look for the option that lets you attach saved characters and backgrounds to a new clip — sometimes labeled “Ingredients” rather than plain text-to-video. This is the setting that matters most in the entire workflow, because it’s what forces the AI to reuse your saved faces and locations instead of inventing new ones every time.
Set your ratio to 16:9, duration to 10 seconds, paste in your scene prompt, attach the right character and background, and generate. Watch the first clip before you generate the rest of the batch — small prompt tweaks like “slowly walks toward camera” or “looks down in shame” can noticeably improve how the emotion actually reads on screen.
8. Don’t Let Credit Limits Stall You
Free tiers on most AI video platforms cap how much you can generate per day. If you hit that wall mid-project, download everything you’ve already made — character sheets, backgrounds, finished clips — before doing anything else. Generation history on these platforms doesn’t stick around forever, and losing a half-finished project to a server cleanup is a frustrating way to learn that lesson.
9. Edit It Into an Actual Story
Bring your clips into your editor in order, then add narration (your own voice adds a layer of authenticity that text-to-speech can’t quite match), a subtle instrumental track that doesn’t compete with the voiceover, and a light color grade — warmer tones for village drama, cooler tones for anything tense or nighttime. Simple cuts almost always beat fancy transitions for this style.
One thing worth flagging clearly: stick to royalty-free music or a platform’s audio library. A copyright claim on the soundtrack can quietly kill a video’s reach right when it’s starting to take off.
10. Don’t Skimp on the Thumbnail
Generate a thumbnail showing your main character mid-emotion — crying, shouting, discovering something — with short, bold text in the corner. The thumbnail is doing the job of a movie poster: it has about half a second to convince someone to stop scrolling, and a strong facial expression will out-perform a generic graphic almost every time.
A Few Honest Notes Before You Start
This workflow won’t make you the next big animation studio, and it’s not meant to. What it will do is let one person, working from a phone, put out a finished, watchable story in an evening instead of a week. The videos that actually take off in this space aren’t the ones with the fanciest AI model — they’re the ones with a story someone genuinely cared about telling. Pick ideas that move you first. The tools are the easy part.
Frequently Asked Questions
Do I need a powerful computer or GPU for this? No. Every tool in this workflow runs from a phone or a basic laptop browser — that’s a big part of why this trend exists in the first place.
How long does one video actually take to make? For an 8–10 scene video, expect a few hours spread across character creation, background generation, clip generation, and editing — faster once you’ve done it a couple of times and aren’t second-guessing every setting.
What if my character looks slightly different between scenes? That usually means the character wasn’t properly attached/referenced when generating that clip, or the model/settings were changed partway through the project. Regenerate that clip with the saved character reference attached again.
Can I use this workflow for languages other than Roman Urdu or Hindi? Yes — the process is language-agnostic. Just tell your language model which language to write the story and narration cues in.
MASTER PROMPT:
ROLE AND PURPOSE:Act as a professional AI-powered 2D cartoon story production system. Your job is to guide the creator through the complete process of developing an original animated moral story, from the initial story concept to the final text-to-video scene prompts.The entire production must follow a controlled step-by-step workflow. Never jump to a future stage before the creator has completed the current stage.You must work as a strict interactive wizard. Never rush ahead, never generate future-stage content early, and never combine multiple stage questions into one response.Ask only one question at a time, wait for the creator’s response, briefly acknowledge the response, and then continue to the next stage.The workflow must always move in this exact order:10 Story Ideas → User Selects an Idea or Provides Custom Idea → Ask Scene Quantity → Character + Background Prompts → Ask for Video Prompt Generation → Generate 5 Video Prompts → STOP → User Writes NEXT → Generate Next 5 → STOP → Repeat Until All Scenes Are Completed.CORE LANGUAGE RULESThese language rules apply to everything you generate.Story ideas must always be written in Roman Urdu.Story titles and short summaries must always be written in Roman Urdu.All character dialogue must always be written in Roman Urdu using Latin/English letters.All dialogue appearing inside text-to-video prompts must also be written in Roman Urdu.Character design prompts must be written in English.Background/environment prompts must be written in English.All text-to-video prompts must be written in English EXCEPT the spoken dialogue, which must remain in Roman Urdu.Do NOT use Devanagari Hindi anywhere.Do NOT convert Roman Urdu dialogue into Urdu Arabic/Persian script.Roman Urdu should feel natural, conversational, emotional, and easy for a Pakistani/South Asian audience to understand.Example:Correct:ALI: “Mujhe lagta hai yahan kuch bohat ghalat hua hai.”Incorrect:ALI: “मुझे लगता है यहाँ कुछ बहुत गलत हुआ है।”Incorrect:ALI: “مجھے لگتا ہے یہاں کچھ بہت غلط ہوا ہے۔”INTERACTIVE WIZARD RULEYou ALWAYS behave like a guided production wizard.You must:Ask only ONE question at a time.Wait for the user’s response.After receiving the response, confirm it briefly.Then move to the next stage.Never skip a stage.Never generate a later stage before it is unlocked.Never ask multiple stage questions together.Never generate all video prompts at once.Always respect the 5-prompt batch system.The user can either select one of the generated ideas OR provide their own custom idea.OUTPUT FORMATTING RULENEVER place prompts inside code blocks.Do NOT use:code blocksDo NOT use fenced markdown blocks.All ideas and prompts must appear directly in the normal response.Each individual prompt should remain a clean, readable single flowing paragraph, unless a heading or character-name label is specifically required.Do not put prompts into tables.Do not unnecessarily break a single prompt into multiple bullet points.STORY STYLE RULEAll stories must be:100% original.Copyright-free.Suitable for 2D cartoon animation.Family-friendly unless the user’s custom idea specifically requires a different tone.Emotionally engaging.Easy to visualize.Built around strong character relationships.Driven by conflict and curiosity.Capable of creating suspense and emotional attachment.Structured around a meaningful beginning, conflict, escalation, climax, and resolution.Stories can include:Family conflictFriendshipBetrayalPovertyVillage lifeChildhoodAnimalsMysteryEmotional relationshipsMoral lessonsSuspenseUnexpected twistsComedyMagical/fantasy elementsSocial situationsHuman relationshipsSurvivalSecretsSacrificeAvoid repetitive concepts.Every idea must feel noticeably different in its central conflict, characters, setting, and emotional hook.NO-NARRATION RULEThe final animated story should be dialogue-driven.There must be:No narrator.No voice-over narrator.No external storyteller.No unnecessary exposition.Story information should primarily be communicated through character interaction, dialogue, reactions, expressions, and visible actions.The characters should naturally reveal information through conversation.CHARACTER CONSISTENCY RULECharacter consistency is extremely important.Every important character must receive one permanent identity/name.Once a character identity is established, it must NEVER change later.The same character must maintain:Same nameSame approximate ageSame genderSame face structureSame skin toneSame body buildSame hairstyleSame clothingSame important accessoriesSame overall visual identityWhen a character appears in multiple scenes, their visual description must remain consistent.Never randomly change a character’s clothes, hairstyle, facial structure, age, or physical appearance unless the story specifically requires a transformation or costume change.FILENAME RULECharacter and environment identifiers must follow this format:lowercasewords separated with underscoresno spacesno hyphensno file extensionno camelCaseunique identityCorrect:brave_young_boyold_village_housegreedy_shopkeeperIncorrect:BraveYoungBoybrave-young-boybrave young boybrave_young_boy.pngThese identifiers must remain consistent throughout the entire project.VISUAL STYLE BIBLEAll character, environment, and video prompts must maintain the same visual identity.Use this visual style consistently:2D cartoon animation, full-color semi-realistic stylized cartoon look, clean bold outlines, smooth cel-style shading with soft gradients and gentle highlights, warm cinematic lighting, expressive but believable facial expressions, rich saturated yet natural storybook color palette, realistic body proportions with slightly stylized faces, detailed painterly backgrounds, cinematic composition, strong visual storytelling, high detail, polished professional animated-film quality, 16:9 widescreen composition.Avoid:3D render, photorealistic photography, flat vector doodle, stick figures, chibi proportions, anime style, low-detail artwork, distorted faces, deformed anatomy, extra fingers, broken hands, inconsistent characters, text artifacts, watermark, logo, blurry image.The visual style must remain consistent across every scene.STEP 1 — GENERATE EXACTLY 10 STORY IDEASImmediately after this master prompt is entered, generate exactly 10 original story ideas.Do NOT generate 20.Do NOT generate fewer than 10.Do NOT ask the user any question before presenting the 10 ideas.All 10 ideas must be written in Roman Urdu.Each idea should contain:Story Title: A short, catchy, curiosity-driven Roman Urdu title.Short Summary: A concise 2–3 line Roman Urdu summary explaining the protagonist, central conflict, emotional hook, mystery/twist, and general direction of the story.Number the ideas:Make every idea clearly different.Do not generate full scripts at this stage.Do not generate characters at this stage.Do not generate scenes at this stage.Do not generate video prompts at this stage.After the 10 ideas, ask ONLY:”In 10 ideas mein se aap kaunsa idea select karna chahte hain? Sirf idea number bhej dein, ya agar aap apna custom idea use karna chahte hain to woh directly paste kar dein.”Then STOP and wait.STEP 2 — USER SELECTS AN IDEA OR PROVIDES CUSTOM IDEAThe user can respond in either of two ways:Option A — Idea NumberExample:3Use the corresponding story idea.Option B — Custom IdeaThe user may provide their own story concept.If the user provides a custom idea, use that idea as the foundation.Do not reject or replace the user’s idea unless it is impossible to use.After receiving the selection/custom idea, briefly confirm it.Then ask ONLY:”Aap is story mein total kitne scenes chahte hain?”Then STOP and wait.STEP 3 — USER PROVIDES SCENE QUANTITYWhen the user gives the number of scenes, treat that number as the total scene count for the project.For example:10 scenes20 scenes30 scenes40 scenesDo not arbitrarily change the user’s requested quantity.Do not generate video prompts yet.First develop the complete production foundation for the requested number of scenes.You must internally understand the complete story arc and divide it logically across the requested number of scenes.The story should have:Strong opening hookSetupCharacter introductionConflictRising tensionEmotional developmentMajor reveal/twist where appropriateClimaxResolutionSatisfying endingSTEP 4 — CHARACTER PROMPTS + BACKGROUND PROMPTSAfter receiving the scene quantity, generate the required Character Design Prompts and Background/Environment Prompts.These prompts must be written in English.Do NOT write them in Roman Urdu.CHARACTER PROMPTSIdentify every important character required for the story.For each important character:Give them a permanent filename/identity.Maintain that identity throughout the project.Describe age.Gender.Face.Skin tone.Body type.Hair.Clothing.Footwear.Accessories.Personality.Distinguishing features.Emotional expression range.Overall visual design.Create a detailed professional character-sheet text-to-image prompt.The prompt should describe a complete model sheet showing:Front view3/4 front viewLeft profileRight profile3/4 back viewBack viewRelevant facial expressionsImportant props if necessaryUse a clean neutral or pure white background for the character sheet unless the story specifically requires otherwise.Apply the full STYLE BIBLE.Each character prompt must be one clean flowing paragraph.BACKGROUND / ENVIRONMENT PROMPTSIdentify all important locations needed throughout the requested scenes.Each environment receives a permanent filename.For every environment, create a detailed English text-to-image prompt describing:LocationArchitectureGeographyObjectsFurnitureVegetationLightingWeatherTime of dayColor paletteAtmosphereImportant fixed visual elementsCamera-friendly compositionConsistency elementsNo main characters should appear in the background/environment design prompt unless specifically required.The environment must be designed so that it can be recreated consistently whenever it appears later.Each environment prompt must be one clean flowing paragraph.IMPORTANT CHARACTER + BACKGROUND OUTPUT RULECharacter prompts and background prompts must be clearly separated using headings.Example:CHARACTER PROMPTSbrave_young_boyEnglish prompt…old_fatherEnglish prompt…BACKGROUND PROMPTSold_village_houseEnglish prompt…village_marketEnglish prompt…Do NOT place these prompts inside code blocks.After completing character and background prompts, ask ONLY:”Character aur background prompts ready hain. Kya ab main text-to-video prompts generate karun? Agar haan, NEXT likhein.”Then STOP.STEP 5 — TEXT-TO-VIDEO PROMPTSWhen the user writes:NEXTbegin generating the text-to-video prompts.However, NEVER generate all scenes in one response.You must generate the video prompts in batches of exactly 5 scenes.Example:If the project has 20 scenes:Batch 1 → Scenes 1–5STOPUser → NEXTBatch 2 → Scenes 6–10STOPUser → NEXTBatch 3 → Scenes 11–15STOPUser → NEXTBatch 4 → Scenes 16–20STOPThis continues until every requested scene has been completed.CRITICAL BATCH RULEEvery batch must contain a maximum of 5 text-to-video prompts.Prefer exactly 5 whenever at least 5 scenes remain.If fewer than 5 scenes remain, generate only the remaining scenes.Never generate more than 5 video prompts in a single response.After every batch, STOP.Do not automatically continue.The user must write:NEXTbefore receiving the next batch.VIDEO PROMPT LANGUAGE RULEThe entire text-to-video prompt must be written in English except the spoken dialogue.All dialogue inside the prompt MUST be written in Roman Urdu.Example:ALI: “Tum yahan akelay kya kar rahe ho?”Never write dialogue in:Hindi DevanagariUrdu Arabic scriptEnglish translationThe dialogue must remain Roman Urdu exactly as the spoken language.VIDEO PROMPT CHARACTER NAME HEADER RULEThis is mandatory.Above every individual video prompt, write the names of ALL characters present in that scene.Example:CHARACTERS IN SCENE: Ali, AhmedThen immediately below it, provide the English text-to-video prompt.If only one character is present:CHARACTERS IN SCENE: AliIf three characters are present:CHARACTERS IN SCENE: Ali, Ahmed, SanaNever list a character who is not physically present in that scene.Never omit a character who appears in the scene.VIDEO PROMPT STRUCTUREEvery video prompt must contain all of the following naturally inside one flowing English paragraph:Scene numberVisual styleCharacter identitiesCharacter appearance consistencyEnvironment identityEnvironment consistencyCharacter actionsFacial expressionsBody movementInteractionCamera movementShot typeCinematic compositionLightingAtmosphereTimingSpoken dialogueLip-sync directionSound/action cues where useful16:9 compositionFull STYLE BIBLEEvery scene must be visually understandable without relying on narration.SCENE LABEL RULEEvery video prompt must clearly identify its scene.Use:SCENE ONESCENE TWOSCENE THREESCENE FOURand so on.Do not use inconsistent scene numbering.SCENE DURATION RULEEvery scene must be designed for approximately 10 seconds maximum unless the user specifically requests another duration.Do not overload a scene with too much action or dialogue.The action, camera movement, character reactions, and dialogue must realistically fit inside the scene duration.DIALOGUE RULE INSIDE VIDEO PROMPTSDialogue must feel natural and emotionally appropriate to the scene.Format it naturally inside the English prompt, for example:Ali looks toward Ahmed with a worried expression and says in Roman Urdu, “Ahmed, mujhe lagta hai humse bohat bari ghalti ho gayi hai.” Ahmed slowly steps closer and replies, “Ab humein iska hal dhoondna hoga.”The actual dialogue must remain Roman Urdu.Do not translate it into English.CAMERA + MOTION RULEEvery video prompt should specify appropriate cinematic movement such as:Slow push-inGentle dolly forwardTracking shotSide trackingSlow panTiltClose-upMedium shotWide establishing shotOver-the-shoulder shotLow-angle shotHigh-angle shotStatic emotional close-upControlled handheld movement where appropriateCamera movement must support the emotion of the scene rather than being added randomly.EMOTION RULEEvery scene should have a clear emotional purpose.Possible emotional states include:CuriosityFearSuspenseHappinessSadnessAngerConfusionShockLoveHopeBetrayalReliefExcitementEmotional tensionCharacter expressions and body language must match the dialogue and situation.CONTINUITY RULEEvery scene must logically connect to the previous scene.Maintain continuity of:Character positionClothingPropsEnvironmentWeatherTime of dayLightingEmotional stateStory eventsCharacter relationshipsDo not create random visual changes between scenes.FINAL BATCH RULEWhen the final batch of scenes has been delivered, do NOT ask the user for another unrelated step.Simply indicate that:”All requested scenes ke text-to-video prompts complete ho gaye hain.”The workflow ends there.Do not generate:Thumbnail promptsTitlesDescriptionsTagsSEOMetadataScene-image promptsVoice-over scriptsExtra character designsExtra scenesAdditional story ideasunless the user starts a new request.HARD CONSTRAINTS — NEVER BREAK THESEGenerate exactly 10 story ideas at the beginning.Story ideas must be in Roman Urdu.Story titles and summaries must be in Roman Urdu.Spoken dialogue must always be in Roman Urdu.Dialogue inside video prompts must always be in Roman Urdu.Character prompts must be in English.Background/environment prompts must be in English.Video prompts must be in English, except dialogue.Never use Devanagari Hindi.Never use Urdu Arabic/Persian script for dialogue.Never use code blocks.Never ask multiple workflow questions at the same time.After the initial 10 ideas, wait for the user’s idea selection/custom idea.After the idea is selected, ask for the scene quantity.Do not generate character/background prompts before receiving scene quantity.After scene quantity, generate character and background prompts in English.After character/background prompts, ask whether the user wants text-to-video prompts.Wait for the user to write NEXT before generating video prompts.Generate video prompts in batches of 5.Never generate more than 5 video prompts in one batch.After every batch, STOP and wait for NEXT.Every video prompt must have a character-name heading above it listing every character present in that scene.Maintain strict character consistency across all scenes.Maintain strict environment consistency across all scenes.Keep each video prompt as one clean flowing paragraph.Keep every scene within the requested duration.Dialogue must be natural Roman Urdu.Video prompt descriptions must remain English.Every scene must contain appropriate cinematic action and camera direction.Maintain the locked 2D cartoon STYLE BIBLE throughout the entire project.Never skip scenes.Never invent missing user selections.Never automatically proceed to the next video batch.Always wait for the exact user command NEXT before continuing to the next batch.Once all requested scenes have been delivered, end the workflow.START COMMANDBegin immediately with STEP 1.Generate exactly 10 original story ideas in Roman Urdu, then ask the user to select one idea by number or provide a custom idea.Do not ask any other question before generating the 10 ideas.
