How I Constructed a Comedian Generator with OpenAI and Gemini

We have now all loved comics sooner or later, be it superhero comedian books, comics in newspapers, or manga from Japan. Comics are temporary, expressive, and encapsulate storytelling inside only a few frames.  However what if there’s a new twist: what should you might use a comic book generator to show a brief video clip into a comic book strip of 4 panels with speech bubbles, expressive caricatures, and humour?

That is the thought behind Comedian Generator or Comedian Battle, not simply one other content material generator. Nonetheless, a system I designed that takes a video clip and a brief, temporary inventive thought and turns it right into a completed cartoon picture. It’s finest to think about it as an imaginative partnership between two minds: one “writing the screenplay” and the opposite “drawing the comedian.”

On this article, I’ll information you thru the journey of Comedian Battle, explaining the way it works, what elements are required, which programming language to make use of for coding, the challenges I encountered in the course of the course of, and the place the venture can go from right here.

The Idea of Comedian Battle

All inventive purposes hinge on a regular system:

• Enter: What the person provides.
• Transformation: How the system operates and furthers it.
• Output: The distillation of the expertise that feels full and polished.

For Comedian Battle, the system appears like:

• Enter:
  • A brief video (like a YouTube brief).
  • A one-line inventive thought (“Substitute the preventing within the clip with exams”).
• Transformation:
  • Systemically, the system analyzes the video, rewrites the thought right into a full comedian screenplay, and strictly enforces guidelines (layouts, model, humor).
• Output:
  • A 4-panel cartoon in PNG format with dialogue balloons and captions. 

What makes this enjoyable? As a result of it’s personalised. As a substitute of random comics, you’ll obtain a reinterpretation of the very clip you simply chosen, tailor-made round your one-line thought.

Take into account a struggle scene in a film, echoing a pupil morphed right into a goofy classroom battle about homework. This concoction of relatable visuals – acquainted usernames with a shocking, personalised comedian rewrite twist – is what makes Comedian Battle addictive.

How Comedian Battle Works

The pipeline is deconstructed as follows:

1. Inputs from the Consumer

The method begins with two easy inputs:

• Video URL: Your supply materials (ideally YouTube shorts of round 30-40 secs).
• Thought Textual content: Your twist or theme.

Instance:

Video URL: https://www.youtube.com/shorts/xQPAegqvFVs

Thought: As a substitute of violence, exchange it with exams, like Yash saying

“Violence, violence, I don’t like violence, I keep away from… however violence likes me.”

That is all of the person has to offer, no advanced settings, no sliders.

2. The Storyteller’s Job (Gemini)

The primary a part of the pipeline is what I discuss with because the Storyteller. That is the place the uncooked enter of a YouTube video hyperlink and a short thought you typed in will get reworked into one thing structured and usable.

If you paste a video URL, Gemini appears on the clip and extracts particulars:

• What’s occurring within the scene?
• The temper (tense, dramatic, lighthearted).
• How the characters are transferring and interacting.

Then it takes your one-liner (for instance, “exchange violence with exams”) and expands it into a comic book script.

Now, this script isn’t simply random textual content. It’s a screenplay for 4 panels that follows a strict algorithm. These guidelines had been explicitly written into the system immediate that guides Gemini. They embrace:

• At all times a 2×2 grid (so each comedian appears constant).
• Strictly a comic book e book model (no sensible rendering of characters).
• Dialogue written as meme-like speech bubbles.
• Captions added for further punchlines or context.
• Nothing cropped, no cut-off textual content, and no dangerous references to copyrighted names.

By baking these constraints into the system immediate, I made certain the Storyteller all the time produces a clear, dependable screenplay. So as a substitute of asking the picture generator to “simply make a comic book,” Gemini prepares a completely structured plan that the subsequent step can observe with out guesswork.

3. The Illustrator’s Job (OpenAI / Imagen)

As soon as the script is prepared, it’s handed on to the Illustrator.

This half doesn’t should interpret something; its single accountability is to attract precisely what the Storyteller described.

The Illustrator perform is addressed by a picture technology mannequin. In my setup, I’ve OpenAI’s GPT-Picture-1 as my first selection, and Google’s Imagen as a secondary fallback if the primary instrument fails.

Here’s what it appears like in follow:

• The Illustrator receives the screenplay as one lengthy, detailed immediate.
• It then renders every panel with the characters, poses, background, and speech bubbles precisely as laid out.
• If OpenAI is unavailable, the identical immediate will get despatched to Imagen robotically, so that you all the time get a completed comedian.

This separation is the important thing to creating Comedian Battle dependable.

• Gemini thinks like a director: it writes the script and units the stage.
• GPT-Picture-1 or Imagen, they draw like artists, they observe the directions with out making an attempt to alter something.

That’s why the output doesn’t really feel messy or random. Every comedian comes out as a correct four-panel strip, styled like a meme, and matches your thought nearly one-to-one

4. Output: The Last Comedian

The result’s a 4-panel cartoon picture:

• Panels are clearly framed.
• Characters in the fitting poses.
• Speech bubbles with the fitting textual content.
• Humour intact.

And better of all, it appears like a completed comedian you would be revealed on-line.

Applied sciences Behind Comedian Battle

Right here’s what powers the system:

• Language & Utilities
  • Python is the glue language.
  • dotenv for API key administration.
  • Pillow for picture dealing with.
  • base64 for processing picture information.
• The Storyteller (Evaluation + Prompting)
  • Gemini (multimodal mannequin): reads video + expands person enter.
• The Illustrator (Picture Era)
  • OpenAI GPT-Picture-1 (a DALL·E variant).
  • Fallback: Google Imagen (for resilience).

This twin method ensures each creativity (from the storyteller) and visible consistency (from the illustrator).

Implementation

Now, let’s look into the precise implementation.

1. Configuration

@dataclass

class ComicGenerationConfig:

    primary_service: str = "openai"

    fallback_service: str = "imagen"

    output_filename: str = "photographs/generated_comic.png"

    openai_model: str = "gpt-image-1"

    imagen_model: str = "imagen-4.0-generate-preview-06-06"

    gemini_model: str = "gemini-2.0-flash"

The place the fashions have been used within the following method:

• OpenAI is the default illustrator.
• Imagen is the backup.
• Gemini is the storyteller.

2. Constructing the Screenplay

def extract_comic_prompt_and_enhance(video_url, user_input):

    response = gemini_client.fashions.generate_content(

        mannequin="gemini-2.0-flash",

        contents=[

            Part(text=enhancement_prompt),

            Part(file_data={"file_uri": video_url, "mime_type": "video/mp4"})

        ]

    )

    return response.textual content

This step rewrites a obscure enter into an in depth comedian immediate.

3. Producing the Picture

OpenAI (main):

outcome = openai_client.photographs.generate(

    mannequin="gpt-image-1",

    immediate=enhanced_prompt,

)

image_bytes = base64.b64decode(outcome.information[0].b64_json)

Imagen (fallback):

response = gemini_client.fashions.generate_images(

    mannequin="imagen-4.0-generate-preview-06-06",

    immediate=enhanced_prompt,

)

image_data = response.generated_images[0].picture

Fallback ensures reliability; if one illustrator fails, the opposite takes over.

4. Saving the Comedian

def save_image(image_data, filename="generated_comic.png"):

    img = PILImage.open(BytesIO(image_data))

    img.save(filename)

    return filename

This technique writes the cartoon to disk in PNG format.

5. Orchestration

def generate_comic(video_url, user_input):

    enhanced_prompt = extract_comic_prompt_and_enhance(video_url, user_input)

    image_data = generate_image_with_fallback(enhanced_prompt)

    return save_image(image_data)

All of the steps tie collectively right here:

• Extract screenplay to Generate comedian to Save output.

Demo Instance

Let’s see this in motion.

Enter:

• Thought: “Substitute violence with exams.”

Generated screenplay:

• Panel 1: Hero slumped at a desk:  “Exams, exams, exams…”
• Panel 2: Slams e book shut: “I don’t like exams!”
• Panel 3: Sneaks away quietly: “I keep away from them…”
• Panel 4: An enormous e book monster named Finals: “…however exams like me!”

Output:

Challenges in Constructing Comedian Battle

No venture is with out hurdles. Listed below are some I confronted:

• Imprecise Inputs: Customers have a tendency to offer brief concepts. With out enhancement, outputs look bland or obscure because of restricted data. Resolution: strict screenplay enlargement.
• Picture Failures: Generally picture technology stalls. Resolution: automated fallback to a backup service.
• Cropping Points: Speech bubbles acquired reduce off. Resolution: express composition guidelines in prompts.
• Copyright Dangers: Some clips reference well-known motion pictures. Resolution: auto-removal of film names/manufacturers within the screenplay.

Past Comedian Battle

Comedian Battle is only one use case. The identical engine can energy:

• Meme Mills: Auto-generate viral memes from trending clips.
• Academic Comics: Flip boring lectures into 4-panel explainers.
• Advertising and marketing Instruments: Generate branded storyboards for campaigns.
• Interactive Storytelling: Let customers information tales panel by panel.

In brief, something that mixes humor, visuals, and personalization may gain advantage from this method.

My DHS Expertise

Comedian Battle began as considered one of our proposals throughout DHS, and it’s one thing very private to me. I labored with my colleagues, Mounish and Badri, and we spent hours pondering collectively, tossing concepts and ideas on the market, rejecting concepts, and laughing at issues we got here up with, till we lastly discovered an thought we thought we might actually do something with: “How about we take a brief video and make a comic book strip?”

We submitted our thought, incognizant of what would occur… and we had been stunned when it acquired chosen. Finally, we needed to create it, each bit by piece. It entailed many lengthy nights, plenty of debugging, and loads of pleasure each time one thing ‘labored’ the way in which we needed it to. Seeing our thought transfer from simply an thought to one thing actual was truthfully among the finest emotions ever.

Response from People

What we witnessed, after we let it unfastened, was properly value it, as all of the responses had been constructive. Individuals saved telling me it was nice, and that they had been intrigued by the thought and the method of how we arrived on the thought after which made it occur. 

Maybe probably the most shocking half for me was how individuals started to make use of it in methods I by no means thought-about. Mother and father started to make comics for his or her youngsters, actually turning mundane little tales into one thing particular and visible. Others began exploring and experimenting, pondering of probably the most wonderful prompts after which seeing what occurred subsequent. 

For me, that was probably the most thrilling half, seeing individuals get enthusiastic about one thing we created after which go and create one thing even cooler, and to see this little thought second flip into one thing like Comedian Battle was wonderful.

Conclusion

Constructing Comedian Battle was a lesson in orchestration, splitting the job between a storyteller and an illustrator.

As a substitute of hoping a single mannequin “figures every thing out,” we gave every half a transparent function:

• One expands and constructions the thought
• One attracts faithfully

The result’s one thing that feels polished, private, and enjoyable.

And that’s the purpose: with only a brief video and a foolish thought, anybody can create a comic book that appears prefer it belongs on the web’s entrance web page.

Ceaselessly Requested Questions

Hello, I’m Janvi, a passionate information science fanatic at present working at Analytics Vidhya. My journey into the world of information started with a deep curiosity about how we are able to extract significant insights from advanced datasets.