How to Bulk‑Create Ghibli‑Style 90s Nostalgia Videos with Free AI Tools

It’s now possible to create a 10–12 minute, Ghibli‑style 90s nostalgia video with consistent characters in just a few clicks using free AI tools. The workflow combines bulk image generation, bulk image‑to‑video conversion, and smart SEO prompts so your final upload has a real shot at going viral on YouTube.

Below is a complete, step‑by‑step breakdown of how to go from idea to fully edited, SEO‑optimized video.

1. Pick a Viral 90s Nostalgia Topic

The entire workflow starts with the right topic. If the topic is weak, no amount of AI magic will save the video.

Use these two rules:

Rule 1: Trigger strong Indian 90s nostalgia. Choose events and memories that instantly transport viewers back in time, such as:

• Eid celebrations in the 1990s before the internet
• Diwali celebrations in the 90s
• Independence Day in school during the 90s
• The craze of Shaktimaan on TV

These topics create an emotional connection, which drives watch time, likes, and shares.

Rule 2: Aim for low competition. Search your idea on YouTube using the keyword plus “Ghibli style”. If you find zero or very few videos, you’ve likely found a strong opportunity. If the topic is saturated, either skip it or create something more unique, more detailed, and longer than existing videos.

If you’re stuck for ideas, you can use a custom “idea generator” prompt in ChatGPT to generate around 15 viral topic ideas at once, then pick your favorite. For example, “The craze of Shaktimaan in the 1990s” is a great candidate.

2. Generate Powerful Nostalgic Scenes with AI

Once you have a topic, the next step is scene creation. This is where most creators cut corners, but it’s the heart of a viral nostalgia video.

Aim for at least 100–200 scenes. Each scene should feel like a real memory, not a scripted ad. The viewer should think, “I’ve done this too.”

Because the final video doesn’t rely on voiceover, each scene must communicate clearly through:

• Action – what the character is doing (e.g., opening a lunchbox, adjusting the TV antenna)
• Body language – how they feel (excitement, shyness, curiosity)
• Objects – context clues (CRT TV, school bag, steel tiffin, ceiling fan)

Use minimal dialogue in your prompts. Only add it when it’s absolutely needed to convey the moment.

If you struggle to come up with so many scenes, use a “scene generation master prompt” in ChatGPT. You can:

• Paste the prompt
• Insert your chosen topic
• Set how many scenes you want (e.g., 200)

ChatGPT will then output scenes in batches (for example, 20 at a time). Keep typing “next” until you have the full list.

3. Design Relatable Characters and Generate Them in Bulk

Viewers remember characters more than individual shots. To make your video memorable, create 4–5 main characters that appear across many scenes.

You can design them yourself by imagining kids, friends, or family from the 90s—how they dress, their hairstyles, and their personalities—or you can use a “character generation master prompt” in ChatGPT to get:

• 4–5 detailed character descriptions
• Ready‑to‑use text‑to‑image prompts for each character

Bulk Character Image Generation with Google Flow + VO Automation

To generate all character images in one click, use Google Flow with the VO Automation Chrome extension:

1. Ask ChatGPT to format all character prompts in a single code block, one per line, like:
CharacterName – full prompt text…

2. Install the “VO Automation – Auto VO on Google Flow” Chrome extension and pin it.

3. In Google Flow, select Text to Image and in the extension:

• Set Concurrent prompts to 1
• Set a random delay between 25–35 seconds (to avoid spam detection)
• Choose model Nano Banana 2
• Set retries to 2 and auto‑download quality to 1K
• Set outputs per prompt to 1
• Pick an output folder

4. Turn off “Ask before every download” in Chrome’s download settings.

5. Paste all character prompts into the extension and hit Run.

The extension will now generate and download all character images automatically.

4. Turn 200 Scene Prompts into Consistent Ghibli‑Style Images

With characters ready, it’s time to create the actual scene images—ideally 100–200 of them in one go.

Create Cinematic Text‑to‑Image Prompts for Each Scene

Each scene image should feel like a mini movie frame. The viewer should instantly understand:

• Where they are
• What is happening
• What the character is feeling

To achieve this, use a “scene‑to‑prompt” master prompt in ChatGPT that:

• Adds characters into each scene
• Describes specific actions (not static poses)
• Emphasizes body language and emotional clarity
• Keeps everything in a consistent Ghibli‑inspired style

Generate prompts in batches (e.g., 20–30 at a time) until you have prompts for all scenes. Then combine them into a single text file.

Before automation, rename each character image file to match the character name used in your prompts (if the name has multiple words, use only the first word). This is crucial for character consistency.

Bulk Scene Image Generation with Image‑to‑Image

Back in Google Flow with the VO Automation extension:

1. Switch to Image to Image in the extension.
2. Set Concurrent prompts to 1 and random delay to 25–35 seconds.
3. Set Maximum input images per prompt to the number of characters (e.g., 5).

In the settings tab:

• Choose Image to Image
• Select model Nano Banana 2
• Set retries to 2 and save

Then:

• Upload all character images
• Enable Auto add character’s image (this is what keeps characters consistent)
• Paste all scene prompts from your text file into the extension
• Verify that each prompt’s serial number matches its scene number
• Choose an output folder and click Run

The extension will generate and download all scene images in sequence. If some fail (marked in red), copy only those failed prompts into a new run, generate them again, then manually rename any mismatched files so that file names match scene numbers (e.g., 1–200).

Sort images by name and confirm you have exactly one image per scene, in order.

5. Generate Image‑to‑Video Clips with Super Grok

Now you’ll turn each scene image into a short video clip using Super Grok’s frame‑to‑video feature, then stitch everything into a full 12‑minute video.

Create Image‑to‑Video Prompts

Use another master prompt in ChatGPT to generate one image‑to‑video prompt per scene. Each prompt should:

• Clearly show what’s happening and why
• Include subtle motion (blinking, cloth movement, fan air, dust particles)
• Use slow, gentle camera moves (push‑in, pan)
• Emphasize strong body language and facial expressions
• Mention ambient sounds and soft music (even if Grok doesn’t handle audio, it guides the visual feel)

Generate these prompts in batches (e.g., 20–30 at a time), then combine them into a single text file, one prompt per line.

Access and Use Super Grok

Once you have access to Super Grok, install the “Grok Automation – Auto Grok on grok.com” Chrome extension and pin it. On grok.com:

1. Open the extension and select Frame to Video.
2. Set Concurrent prompts to 1.
3. Set a random delay between 20–40 seconds.

In the settings tab:

• Choose Frame to Video
• Set aspect ratio to 16:9
• Choose video duration (around 6 seconds works well for this workflow)
• Set retries to 2
• Set resolution to 720p and save

Back in the control section:

• Upload all scene images, sorted by name in ascending order so they stay in sequence
• Ensure “Use start frame only for each prompt” is selected
• Paste all image‑to‑video prompts into the prompt box
• Verify that image number 1 matches prompt 1, image 2 matches prompt 2, and so on

Choose an output folder and click Run. The extension will generate and download all video clips automatically.

6. Edit, Enhance, and Remove Watermarks

Import all the generated clips into your video editor (Premiere Pro, DaVinci Resolve, CapCut, etc.) and place them on the timeline in order.

Remove the Grok watermark:

• Select all clips
• Scale them to about 118%
• Adjust the Y position (e.g., -28) until the watermark is cropped out

Then:

• Trim unwanted parts from each clip to improve pacing and storytelling
• Add an “old film” or similar effect to enhance the nostalgic feel
• Optionally add simple text overlays to explain context or emotions in key scenes

This manual editing is important for monetization and originality. Platforms like YouTube want to see human input, not raw, unedited AI output.

When you’re happy with the edit, export in 4K (3840×2160) at high quality.

7. Optimize for YouTube SEO: File Name, Title, Tags, Description

A strong video alone doesn’t guarantee views. You also need smart metadata so the algorithm understands and surfaces your content.

Rename the Video File with SEO in Mind

Before uploading, rename your exported file with a clean, keyword‑rich name that matches your topic. This gives YouTube an early signal about your content.

You can use an “enhanced file name” master prompt in ChatGPT to generate 5 SEO‑optimized file name ideas based on your topic, then pick one and rename your file.

Create a Viral, High‑CTR Title

Use an SEO title generator prompt in ChatGPT to:

• Generate around 10 title options
• Include strong keywords (e.g., “90s India”, “Ghibli style”, “nostalgia”, “Shaktimaan”)
• Maximize click‑through rate with emotional hooks

Copy the best suggested title and use it as your YouTube title.

Generate Tags and Description

Next, use a tag generation master prompt to create a list of SEO‑optimized tags based on your title and topic. Paste these into the YouTube tags field.

Then use a description master prompt to generate a detailed, keyword‑rich description that:

• Summarizes the video
• Naturally includes your main tags and keywords
• Reinforces the nostalgic theme and Ghibli‑style visuals

Paste this into the description field.

8. Design High‑CTR Thumbnails with AI

Thumbnails can make or break your video’s performance. To generate them with AI:

1. Use a thumbnail master prompt in ChatGPT.
2. Provide 2–3 short text options you might want on the thumbnail (comma‑separated).
3. Insert your video topic.

The prompt will output three detailed text‑to‑image thumbnail prompts plus suggested file names. Use these prompts in Google Flow (Text to Image) to generate three thumbnail variations in 2K quality.

Download them, rename using the suggested file names, and upload your favorite as the main thumbnail. You can also A/B test multiple thumbnails over time.

For more advanced, free video generation workflows, you may also want to explore setups like running LTX 2.3 for local text‑to‑video on a cloud GPU, which can complement this image‑to‑video pipeline.

9. Final YouTube Settings and Publishing Strategy

Before publishing:

• Add the video to a relevant playlist (e.g., “90s Nostalgia Ghibli Style”)
• Mark it as altered content since it uses AI‑generated visuals
• Add end screens and cards to keep viewers on your channel longer
• Schedule the video for a time when your target audience (for example, Indian viewers) is most active, often in the evening

If your channel is small (under 1,000 subscribers), don’t be discouraged if the video doesn’t blow up instantly. It can take 1–2 weeks for YouTube to fully test and push your content.

By combining strong nostalgic topics, carefully engineered prompts, bulk AI automation, and solid SEO, you can consistently produce long, cinematic Ghibli‑style videos that stand out—and you’ll be ready to plug into future AI video advances as new models and tools roll out, whether from Grok or other emerging platforms like those covered in deep‑dive pieces on next‑generation models such as Claude Mythos.