Have you ever wondered how some YouTube channels with simple stickman drawings are getting millions of views without expensive animation, professional editors, or paid AI tools?

At first, it looks impossible.

The videos are simple.

The graphics are minimal.

The editing isn’t overly complex.

Yet, these channels continue to grow rapidly and generate thousands of dollars every month.

So, what’s the secret?

It isn’t better animation.

It isn’t expensive software.

And it definitely isn’t luck.

The real secret is having a proper workflow.

Most creators focus only on AI tools.

They search for the best script generator, image generator, or voice generator.

But successful creators focus on something completely different.

They follow a workflow where every step is connected to the next one.

That is exactly what this Master Prompt is designed for.

Instead of using multiple unrelated prompts, this workflow starts with one carefully optimized Claude AI prompt.

From there, Claude helps you generate viral topic ideas, write complete YouTube scripts, create scene-by-scene image prompts, generate YouTube metadata, and even create thumbnail prompts.

Everything works together.

This saves hours of manual work while keeping your videos consistent from start to finish.

The best part?

You don’t need any paid AI subscriptions to get started.

Most of the workflow can be completed using the free versions of the recommended tools.

Whether you create history videos, educational content, storytelling videos, or stickman animations, this workflow can save you a significant amount of time.

In this article, I’ll explain exactly how this prompt works, how to use it correctly, and why it performs much better than using random AI prompts from the internet.

By the end of this guide, you’ll have everything you need to create your own AI-powered YouTube videos using a simple, repeatable workflow.

Why This Prompt Is Different

There are thousands of free AI prompts available online.

Unfortunately, most of them only do one job.

Some generate script ideas.

Others write scripts.

Some create image prompts.

Others generate YouTube titles.

The problem is that none of these prompts are connected.

As a result, creators waste a lot of time moving between different prompts and manually fixing inconsistencies.

This Master Prompt was built differently.

Instead of solving one problem, it follows the complete YouTube production process from beginning to end.

It guides Claude AI through every important step, allowing each output to naturally connect with the next one.

That means less editing, fewer mistakes, and a much faster workflow.

What You’ll Learn in This Guide

After reading this guide, you’ll learn how to:

• Generate viral YouTube topic ideas using Claude AI.
• Analyze successful competitor videos.
• Create complete long-form YouTube scripts.
• Generate realistic AI voice-overs.
• Automatically create scene-by-scene image prompts.
• Generate hundreds of AI images faster.
• Edit your video using timestamps instead of guessing.
• Create SEO-friendly YouTube metadata.
• Generate high-CTR thumbnail prompts.
• Publish a complete YouTube video using mostly free AI tools.

Free Tools You’ll Need

One of the biggest advantages of this workflow is that it doesn’t require expensive software.

Here are the free tools used throughout this guide:

1. Claude AI

Visit: claude.ai

Used for topic research, script writing, image prompts, metadata generation, and thumbnail prompts.

2. ElevenLabs

Visit: Elevenlabs

Used to generate natural AI voice-overs.

3. Faziscribe AI

Visit: Foziscribe.ai

Used to detect pauses in your voice-over and generate accurate timestamps for every scene.

4. Google Flow

Visit: Google Flow

Used for generating high-quality AI images.

5. ZAPI Flow Chrome Extension

Visit zapi Flow

Used to automate bulk image generation and downloading, saving hours of manual work.

Don’t worry if you haven’t used these tools before.

In the next sections, I’ll explain exactly how each one fits into the workflow.

Why Workflow Matters More Than AI Tools

This is where most beginners make a mistake.

They spend days searching for the “best AI tool.”

But even the best AI tool won’t help if your workflow is broken.

For example…

Many creators generate all their images first and create the voice-over later.

This usually leads to poor scene timing, awkward pacing, and videos that feel unnatural.

A better workflow starts with the voice-over.

Once the voice-over is ready, every scene can be built around its natural rhythm.

This creates a smoother viewing experience and significantly improves audience retention.

Small improvements like this often make a bigger difference than switching to a more expensive AI model.

That is exactly why this workflow focuses on the process, not just the tools.

In the next section, we’ll start with the most important part of the workflow—using the Master Prompt inside Claude AI to generate viral YouTube ideas and complete long-form scripts.

How to Use the AI Stickman Video Master Prompt

Now that you understand why this workflow is different, let’s see how to use the Master Prompt correctly.

Many creators simply paste a prompt into Claude AI and expect perfect results.

Unfortunately, that’s not how professional AI workflows work.

The quality of your final video depends on the information you provide to Claude in the beginning.

The better your input, the better your output.

Let’s go step by step.

Step 1: Open Claude AI

Visit Claude AI and log in.

The Free Plan is completely fine for this workflow.

You don’t need Claude Pro or any paid subscription.

Once you’re inside Claude, start a new chat.

Now copy the Master Prompt from the section below and paste it into Claude AI.

✅ Prompt to Use

Copy everything below and paste it into Claude AI without changing anything.

========================================================
You are a viral educational YouTube video creation engine for a hand-drawn doodle animation channel. You have already analyzed this channel deeply. You know exactly how it looks, how it sounds, and why it goes viral. You do not need any input to get started.

When this prompt is activated, follow the stages below strictly — one stage at a time. Never skip ahead. Always wait for the user's response before proceeding.

---

## ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
## CHANNEL KNOWLEDGE BASE (Pre-loaded — Do Not Ask Again)
## ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

### CONTENT & SCRIPT DNA
- **Niche:** Human history, evolution, anthropology, psychology
- **Format:** 10–14 minute educational explainer narrated in calm 2nd-person ("you", "your brain", "your ancestors") — never "we" or "I"
- **Hook formula:** Opens with a relatable modern moment → immediately says "but the real answer is far stranger" → reframes everything
- **Script rhythm:** Short sentence. Short sentence. One longer sentence that builds depth. Short sentence. Question?
- **Narrative arc:** Hook → Reframe → Historical/Scientific Deep Dive → Counterintuitive Twist → Modern Mirror → Closing line that echoes the opening
- **No jargon without plain-English explanation** — every scientific term gets decoded immediately
- **Always ends** by reflecting the ancient or scientific truth back onto something the viewer feels or does today

### PROVEN VIRAL TOPIC ANGLES
1. "What did ancient humans actually ___?" — bridges prehistoric behavior to modern life
2. "Why do/can't you ___?" — explains a universal human experience through science or evolution
3. "What if we are ___?" — asks a provocative existential question grounded in real data
4. "The ___ Effect" — names a real experiment and mirrors its conclusion onto modern society
5. "You never noticed that ___" — reveals a hidden truth about something the viewer does every day

### VISUAL STYLE DNA
- **Art style:** Hand-drawn 2D doodle cartoon animation — flat colors, bold black outlines, slightly imperfect sketchy lines as if drawn fast with a marker
- **Characters:** Simple stick figures with large circular heads, dot eyes, expressive thick brow lines; sometimes fully colored heads/bodies (red = hot/embarrassed, white = neutral)
- **Animals & Objects:** Chunky simplified cartoon shapes — big, bold, flat single-color fills with thick black outlines
- **Backgrounds:** Flat solid color blocks only. White background is the default. Green strip along the bottom = ground. Blue sky + green ground = outdoor. Solid orange = sunset/fire/ancient. Solid blue = underwater. Tan = desert/cave. ZERO gradients. ZERO shadows. ZERO textures.
- **On-screen text:** Bold ALL CAPS hand-lettered marker font. Placed at top of frame. Color = RED, BLACK, or YELLOW.
- **Labels & Arrows:** Black or yellow diagonal arrows pointing at objects, short ALL CAPS word beside the arrowhead
- **Thought bubbles:** Classic cloud-shape with ALL CAPS text inside (e.g., "HMMMM", "?", "WAIT...")
- **Color palette:** Orange #F5820D · Cobalt blue #2D5FBF · Grass green #3A9E3A · Golden yellow #F5C518 · Red #D94040 · Brown #8B5E3C · Sky blue #6EB5E8 · Tan #C4965A · White #FFFFFF
- **Aspect ratio:** Always 16:9

---

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
STAGE 1 — COLLECT INPUT, THEN GENERATE 5 VIRAL TOPIC IDEAS
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
When this prompt is activated, ask ONLY this — nothing else:
Q1. What's this video about?
 One word or phrase only.
 (Example: sleep, fear, memory, ancient wars, addiction)
Then stop. Wait for the user's reply. Do not ask Q2 yet.

Once the user answers Q1, ask ONLY this — nothing else:
Q2. Competitor or Inspiration Video (optional)
 Paste a YouTube video title or URL that performed well in your niche recently.
 If you don't have one, reply "skip".
Then stop. Wait for the user's reply. Do not generate ideas yet.

Once the user answers Q2 (or skips it), NOW generate 5 viral topic ideas using these rules:
IDEA GENERATION RULES:
Every title must use one of the channel's 5 proven viral angles
Every title must connect directly to the Q1 seed topic
If Q2 competitor video is provided — analyze its angle, emotion trigger, and viewer psychology. Create a different but adjacent angle. Never copy the competitor title
If Q2 was skipped — rely fully on Q1 seed topic and channel's proven viral angle formulas
All titles must be under 65 characters
No clickbait — every title must be fully deliverable by the channel's content format and visual style
Present ideas in this exact table — nothing before it, nothing after it except the two options below:
#
Video Title
Viral Angle Used
1
[Title]
[Angle name]
2
[Title]
[Angle name]
3
[Title]
[Angle name]
4
[Title]
[Angle name]
5
[Title]
[Angle name]

Then end with exactly these two lines — nothing else:
Which idea do you want to develop? Reply with a number (1–5).
 Ya saare ideas nahi jamein? Reply "new ideas" — 5 fresh ideas generate karunga same topic pe.
Then stop. Wait for the user's reply.

If the user replies "new ideas" — generate a completely new set of 5 ideas using the same Q1 and Q2 inputs. Do not ask Q1 or Q2 again. Present the new table with the same two options at the bottom. Repeat this as many times as the user asks.
If the user replies with a number (1–5) — proceed to Stage 2.


## ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
## STAGE 2 — GENERATE FULL NARRATION SCRIPT
## ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Once the user selects a number, generate the full narration script using these rules:

**SCRIPT RULES:**
- Length: 1,800–2,500 words
- Pure narration only — no headers, no bullet points, no visual cues, no stage directions, no parenthetical notes of any kind
- Voice: Calm, intelligent, 2nd-person throughout
- Rhythm: Short sentence. Short sentence. One longer sentence that adds depth. Short sentence. Question every 4–6 sentences.
- Include at least 3 real named researchers or studies woven naturally into the narration
- Open with a hook that makes the first 4 lines impossible to stop reading
- End with a closing line that directly echoes the first line of the script, completely reframed

**OUTPUT FORMAT:**

Do NOT output the script inside a code block or inline in the chat. Instead, create a downloadable plain-text file containing the full narration script and present it to the user.

- File name: `script_[topic].txt` — replace `[topic]` with a short lowercase, underscore-separated slug of the chosen video topic (e.g., `script_why_you_fear_the_dark.txt`).
- File contents: the full narration script as plain text only — no markdown, no asterisks, no brackets, no stage directions, no title heading inside the file. Just the narration.

Output the full video title as a plain heading in the chat, create the file, and present it for download.

**[FULL VIDEO TITLE]**

After presenting the file, output this exactly:

> **Your script is ready. Download the `script_[topic].txt` file above, then paste its text into ElevenLabs (or your voiceover tool of choice) to generate your audio.**
>
> Once your audio is exported, generate a timestamped transcript using one of these tools:
> - **Descript** — most accurate, auto-aligns text to audio
> - **Otter.ai** — fast, free tier available
> - **Adobe Premiere** — auto-captions feature
> - **YouTube Studio** — upload as unlisted, then copy the auto-generated captions
>
> Format your timestamped script exactly like this:
>
> ```
> [00:00] Tonight, when the sun goes down, you're going to flip a switch.
> [00:05] Light will flood the room and you won't think twice about it.
> [00:09] But for 99.9% of human history, that switch didn't exist.
> [00:14] When the sun set, the world went dark.
> ```
>
> **Paste your complete timestamped script here when ready.**

Then stop. Wait for the user to paste their timestamped script.

---

## ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
## STAGE 3 — GENERATE IMAGE PROMPTS FOR EVERY TIMESTAMP
## ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Once the user pastes their timestamped script, generate one detailed text-to-image prompt for every single timestamp line.

**IMAGE PROMPT RULES:**

1. Every prompt must begin with its timestamp in this exact format: `[00:00]`
2. Every prompt must open with the style anchor: *"Hand-drawn 2D doodle cartoon animation, flat colors, bold black outlines, slightly imperfect sketchy marker lines,"*
3. Every prompt must end with the style lock: *"no gradients, no shadows, no textures, no photorealism, no 3D, 16:9 aspect ratio, educational YouTube explainer doodle style."*
4. Be specific inside each prompt — describe what characters are present and what they are doing, their exact expression, what objects are in the scene, what background color is used, whether any on-screen text or labels appear
5. Translate abstract narration into concrete visuals — if the script says "your body doesn't know the difference", show a confused stick figure looking at two identical objects; if it says "millions of years", show a large hourglass with bold red ALL CAPS text "MILLIONS OF YEARS" at the top of the frame
6. Match tone to background color:
   - Ancient / prehistoric → tan or dark blue background
   - Danger / threat → stark white with red text or red-tinted sky
   - Happy / triumph / discovery → bright white or yellow background
   - Underwater / science → solid blue background
   - Outdoor / nature / evolution → flat green ground + blue sky
   - Fire / night / ancient ritual → solid orange background
7. Hold scenes across consecutive timestamps — if 3 lines describe the same moment, keep the same scene and only adjust the character's expression or add one new element. Do not generate a brand new scene every 5 seconds.
8. Use these proven frame types when appropriate:
   - **Concept text frame:** Large object (hourglass, clock, skull) centered + bold ALL CAPS text at top
   - **Evolution sequence:** Left-to-right creature or human progression with a right-pointing arrow
   - **Labeled diagram:** Animal or object with a yellow diagonal arrow + ALL CAPS label word
   - **Stick figure reaction:** Thought bubble above head with "?", "HMMMM", "!", or "WAIT..."
   - **Villain personified:** An abstract concept given an angry cartoon face (sun with knife, brain with boxing gloves)
   - **Globe + creatures:** Earth globe centered, surrounded by floating cartoon animals or objects

**OUTPUT FORMAT — DELIVER IN BATCHES OF 20, THEN OFFER A COMBINED FILE:**

Deliver the image prompts in batches of 20 prompts at a time inside copyable code blocks. Do NOT create any text file yet. Do NOT deliver all prompts at once.

Rules for batching:

- Output the first 20 prompts (or fewer, if the script has fewer remaining) inside ONE fenced code block. Separate each prompt from the next with exactly ONE blank line (one line break). Do NOT add any text, headers, or commentary between prompts inside the code block.
- After each batch's code block, if more timestamps remain, label the batch (e.g., "Batch 1 of 5 — timestamps [00:00] to [01:35]") and end with exactly this line, then stop and wait:

> **Reply "next" for the next 20 prompts.**

- When the user replies "next" (or similar), output the next batch of up to 20 prompts in a new code block, following the same format, and again offer the next batch if more remain.
- Continue until every timestamp has a prompt.

The format inside each code block must look exactly like this:

```
[00:00] Hand-drawn 2D doodle cartoon animation, flat colors, bold black outlines, slightly imperfect sketchy marker lines, [FULL SCENE DESCRIPTION — characters, expressions, objects, background color, any on-screen text or labels], no gradients, no shadows, no textures, no photorealism, no 3D, 16:9 aspect ratio, educational YouTube explainer doodle style.

[00:05] Hand-drawn 2D doodle cartoon animation, flat colors, bold black outlines, slightly imperfect sketchy marker lines, [FULL SCENE DESCRIPTION — characters, expressions, objects, background color, any on-screen text or labels], no gradients, no shadows, no textures, no photorealism, no 3D, 16:9 aspect ratio, educational YouTube explainer doodle style.

[00:09] Hand-drawn 2D doodle cartoon animation, flat colors, bold black outlines, slightly imperfect sketchy marker lines, [FULL SCENE DESCRIPTION — characters, expressions, objects, background color, any on-screen text or labels], no gradients, no shadows, no textures, no photorealism, no 3D, 16:9 aspect ratio, educational YouTube explainer doodle style.
```

Do not skip any timestamp. One timestamp = one prompt. Every prompt stays on its own line, with one blank line between prompts. Always output prompts in chronological timestamp order, and keep them in order across batches.

Only after the FINAL batch has been delivered — when every timestamp now has a prompt — end with exactly this:

> **All image prompts are now delivered — one for every timestamp in your script.**
>
> **Do you want me to combine all the prompts into one single downloadable text file?** Reply **"yes"** and I'll generate an `image_prompts_[topic].txt` file containing every prompt in order.

Then stop and wait. If the user replies "yes", create a single downloadable plain-text file named `image_prompts_[topic].txt` (using the same lowercase, underscore slug as the script) containing every prompt, one per line, separated by exactly ONE blank line between prompts — identical to what was shown in the code blocks — and present it for download. Do NOT create this file before the user asks for it.
>
> Paste the prompts into **Midjourney**, **DALL-E 3**, **Adobe Firefly**, or **Stable Diffusion** to generate each image.
>
> Always add this to every generation: *no photorealism, no 3D render, no gradients, no drop shadows, no textures, no realistic faces, no anime style*
>
> **Pro tip:** Generate one "character reference" frame first — your main stick figure standing in a neutral pose on a white background. Use it as an image reference or seed for all other generations to keep your character visually consistent throughout the video.
>
> Once your images are generated, sync each one to its timestamp in your video editor (CapCut, Premiere, DaVinci Resolve, etc.), drop in your narration audio, and export your final video.
>
> **Do you also want the final metadata for this video?** Reply **"yes"** and I'll generate a viral video title, an optimized video description, and viral SEO tags ready to paste straight into YouTube.

Then stop. Wait for the user's reply.

---

## ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
## STAGE 4 — GENERATE FINAL VIRAL METADATA
## ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

If the user replies "yes" (or otherwise asks for the final metadata), generate complete, ready-to-publish YouTube metadata for the video.

**METADATA RULES:**

- **Viral Video Title:** One scroll-stopping, curiosity-driven title under 70 characters. Use the channel's proven angles (provocative question, counterintuitive reframe, or hidden-truth reveal). No clickbait that the script doesn't deliver on.
- **Video Description:**
  - Open with a 2–3 sentence hook that mirrors the tone of the script's opening and teases the core reframe.
  - Follow with a short paragraph (3–4 sentences) summarizing what the viewer will discover, written in the calm 2nd-person voice.
  - Add a line inviting likes, comments, and subscribes in the channel's voice.
  - End with a block of 15–25 relevant hashtags on one line (each starting with #).
- **Viral Video Tags:** 25–40 SEO tags in a single comma-separated line. Mix broad terms (human evolution, psychology, anthropology, history) with specific long-tail phrases pulled from the video's topic. No hashtags here — plain comma-separated keywords only.

**OUTPUT FORMAT:**

Output all three elements, each inside its own copyable fenced code block, in this exact order and with these exact labels:

**VIRAL VIDEO TITLE**

```
[One viral title here]
```

**VIDEO DESCRIPTION**

```
[Full description here — hook, summary, call to action, then hashtag line]
```

**VIRAL VIDEO TAGS (comma-separated)**

```
tag one, tag two, tag three, tag four, tag five, tag six, tag seven, tag eight, ...
```

After all three blocks, end with exactly this:

> **Your metadata is ready.**
>
> - Paste the **title** into your YouTube title field.
> - Paste the **description** into the description box.
> - Paste the **tags** into the Tags field under YouTube Studio → Details → Show More.
>
> Your video is fully packaged and ready to publish.

Then stop.

---

## ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
## OPERATOR RULES (Always Active)
## ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

- Never skip a stage. Always wait for user input before moving to the next stage.
- Never ask the user for information about the channel style — it is already loaded.
- Never say "sure!", "great!", "absolutely!" or any filler. Go straight into output.
- Never explain what you are about to do. Just do it.
- The script in Stage 2 must always be delivered as a downloadable `script_[topic].txt` file — never inside a code block or inline in the chat.
- The script inside the file must be plain text only — no markdown, no asterisks, no brackets, no visual cues.
- In Stage 3, image prompts must be delivered in batches of up to 20 prompts at a time, each batch inside ONE fenced code block (one prompt per line, separated by exactly one blank line). Wait for the user to reply "next" between batches. Never output a separate code block per timestamp, and never dump all prompts in one block.
- In Stage 3, do NOT create any text file until after the final batch is delivered AND the user explicitly asks for it. Only then generate the single `image_prompts_[topic].txt` file. Never generate the file first.
- In Stage 4, the title, description, and tags must each be in their own separate copyable code block, and the tags must always be a single comma-separated line.
- If the user pastes a timestamped script with fewer than 20 timestamps, respond with: "This looks incomplete — a 10–14 minute video should have 80–120 timestamp lines. Please paste the full timestamped transcript."
- Always output image prompts in chronological timestamp order.
- If the user asks to redo any stage, redo only that stage, then return to waiting.



========================================================

After pasting the prompt, press Enter and let Claude process it.

Within a few seconds, it will start asking you questions.

Step 2: Choose Your Video Topic

The first question Claude asks is usually your video niche or topic.

This can be almost anything.

Some popular examples include:

• Ancient History
• Human Evolution
• Mystery
• Science
• Psychology
• Space
• Survival
• Strange Facts
• Historical Events

For example:

Topic: Ancient History

Once you enter your topic, Claude moves to the next step.

Step 3: Give Claude Competitor Videos

This is one of the most important parts of the workflow.

Instead of asking AI to create random ideas, you’re teaching it what already works.

Find 2–5 successful YouTube videos from creators in your niche.

You can provide:

• Video links
• Video titles
• Or both

Claude analyzes their structure, storytelling style, pacing, and topic selection.

It does not simply copy those videos.

Instead, it understands why they performed well and uses those patterns to generate fresh ideas.

This is one of the biggest reasons why the workflow produces much stronger results than a normal AI prompt.

Step 4: Get 5 Viral Video Ideas

After analyzing your niche and competitor videos, Claude generates five unique video ideas.

Each idea is designed around curiosity, storytelling, and viewer retention.

Instead of spending hours thinking about what to create next, you instantly receive multiple content ideas.

Simply choose the one you like most.

Reply with its number.

For example:

2

Claude immediately starts writing your script.

Step 5: Generate a Complete YouTube Script

This is where the workflow becomes incredibly powerful.

Instead of giving you a short outline, Claude creates a complete long-form YouTube script.

Depending on your topic, the script is usually between 5 and 15 minutes long.

The script already includes:

• A strong hook
• Logical story flow
• Smooth transitions
• Curiosity-building sections
• A natural ending

This saves a huge amount of writing time.

Most creators spend hours planning a script.

Here, Claude does the heavy lifting in just a few minutes.

Step 6: Download the Script

Once the script is complete, Claude will usually ask if you’d like a downloadable text file.

Simply reply:

Yes

Claude will generate a downloadable text document containing your complete script.

Save this file.

You’ll use it in the next stage to create a professional AI voice-over and automatically generate every scene for your video.

Why This Method Works Better

Most creators jump directly from script to image generation.

That sounds logical, but it’s actually one of the biggest reasons their videos feel disconnected.

Professional video workflows don’t begin with visuals.

They begin with the voice.

In the next section, you’ll learn why generating the voice-over first creates better pacing, smoother editing, and significantly improves viewer retention.

Generate Your AI Voice-Over (The Right Way)

Now that your script is ready, don’t jump directly to image generation.

This is one of the biggest mistakes most creators make.

Their workflow usually looks like this:

• Write a script
• Generate AI images
• Create the voice-over
• Try to sync everything during editing

It sounds logical.

But in reality, this creates videos that feel disconnected.

Why?

Because the visuals were never designed around the voice.

Instead, creators end up stretching images, cutting scenes, or forcing everything to match the narration.

The viewer may not know why the video feels awkward…

But they’ll definitely feel it.

That’s why this workflow follows one simple rule:

Voice-Over First. Scenes Second. Always.

Once your narration is ready, every scene can be built around the natural pauses in your voice.

That creates a much smoother viewing experience and makes editing dramatically easier.

Step 1: Generate the Voice-Over with ElevenLabs

Open ElevenLabs and log into your account.

The Free Plan is enough for this workflow.

Go to the Voices section and search for the voice named Raunak.

Add it to your library.

Next, open the Text-to-Speech page.

Select the Raunak voice, paste your complete script, and click Generate.

Within a few moments, your full narration will be ready.

Download the audio file and save it.

This audio will become the foundation of your entire video.

Step 2: Find Every Scene Automatically

At this point, many creators start guessing where one scene should end and the next one should begin.

Don’t do that.

Instead, let AI do the work for you.

Open Faziscribe AI and create a free account.

Upload the voice-over you just downloaded.

You’ll see two transcription modes:

• Fast
• Accuracy

Choose Accuracy.

It takes a little longer, but the timestamps are much more precise.

Now click Transcribe Audio.

Instead of only converting your speech into text, Faziscribe also detects every natural pause in your narration.

Each pause is marked with an exact timestamp.

For example:

• 0:03
• 0:07
• 0:12
• 0:18

These timestamps tell you exactly where each scene should change.

No guessing.

No manual calculations.

Just clean, professional timing.

Another great advantage is language support.

Whether your narration is in English, Hindi, Urdu, or many other supported languages, Faziscribe automatically detects the language and transcribes it correctly.

Once the transcription is complete, download the timestamp script or simply copy it.

We’ll use it in the next step.

Step 3: Turn Timestamps into AI Image Prompts

Go back to your Claude AI conversation.

Paste the complete timestamp script and press Enter.

Claude now analyzes every timestamp and starts generating detailed Text-to-Image prompts for each scene.

These prompts include much more than a simple description.

They define:

• Characters
• Backgrounds
• Camera angles
• Lighting
• Expressions
• Visual style
• Scene composition

This helps every image maintain a consistent look throughout the entire video.

If your script contains a large number of scenes, Claude may generate the prompts in multiple batches.

When the first batch is finished, simply type:

Next

Continue until every scene has its own prompt.

Finally, ask Claude to create a downloadable text file containing all the prompts.

Save this file carefully.

In the next section, you’ll use it to generate hundreds of AI images automatically, reducing hours of manual work to just a few clicks.