The following tips will help give you a competitive advantage with the latest version of ChatGPT:
→ Capture Your Writing Style
Feed GPT a few samples of your writing and ask it to create a style guide for future outputs.
Example prompt:
"Analyze the text below for style, voice, and tone. Using NLP, create a prompt to write a new article in the same style, voice, and tone: [insert your text or link to your writing]"
→ Content With A Purpose
Generate content with a specific goal in mind, telling GPT who your audience is, and what you want to achieve with your content.
Example prompt:
"Topic: How to grow your email list
For audience: Social media users.
Content goal: Motivate audience to feel excited about growing their email list while teaching them one tip."
→ Expert Simulator
Ask GPT to play the role of a customer or skilled expert, and generate content as if it were that specific persona.
Example prompt:
"Act as a seasoned data scientist and guide me in writing a report that highlights the latest trends and predictions in the field of artificial intelligence. Provide me with unique insights and expert advice that will help me deliver high-quality, informative content to my audience."
→ Unconventional Prompts
Try using prompts that are more open-ended or abstract to get unique and creative responses. You can unlock GPT's creative potential in finding vivid language and unexpected topics.
Example prompt:
"Write a poem about content creation" or "Describe feeling like an digital marketer in 10 adjectives."
→ Ultra-Brainstorming
Instead of asking GPT to generate a list of potential topic ideas for your next project or ask for new angles to cover a familiar topic.
Example prompt:
"Topic: Improving your copywriting skills.
For the above topic, come up with new and innovative ideas that are out-of-the-box and unconventional."
→ Challenge Mainstream Thought
Ask for examples that contradict the dominant narrative to seek out provocative angles that defy expectations and break the mold.
Example prompt:
"Give examples that contradict the ideas expressed on this webpage, and generate thought-provoking solutions: [insert webpage URL]"
→ Experiment With Styles & Tone
Write in different styles or tones to create more dynamic and varied content.
Style examples:
Descriptive
Persuasive
Narrative
Expository
Technical
Academic
Creative
Informal
Formal
Humorous
Satirical
Ironical
Dramatic
Poetic
Mystical
Suspenseful
Romantic
Realistic
Surrealistic
Gothic
Tone Examples:
Formal
Informal
Serious
Humorous
Sarcastic
Ironic
Cynical
Satirical
Confident
Doubtful
Optimistic
Pessimistic
Sympathetic
Empathetic
Angry
Aggressive
Respectful
Playful
Sincere
Objective
Example prompt:
"Give [insert style] advice you can about using GPT-4 to create more effective content in a [insert tone]."
→ Different Perspective
Ask it to write from the perspective of a group of characters with different backgrounds.
Example prompt:
"Topic: Productivity for entrepreneurs.
For the above topic, write multiple perspectives from a group with different viewpoints using their own voice and phrasing."
→ Format Switch
Use GPT to write in different formats, such as outlines, mind maps, bullet points, persuasive essays, or chunks of text of less than 280 characters, using a specific structure.
Example prompt:
"Create a mind map on the topic of using Notion to stay organized as a content creator, listing out the central idea, main branches, and sub-branches."
→ Human-Written Inspiration
Ask GPT to work according to information you've read elsewhere, like persuasive writing tips from a blog post.
Example prompt:
"Write a brief post about why copywriting is an essential skill in 2023. Use the insights from the following webpage: [insert webpage URL]"
Pro Tip: GPT-4 can now process web links and images - use this to your advantage when prompting [or ask GPT the best way to do so].
Hyperparameters are adjustable settings that determine the behavior of the language model during training and inference.
Changing these hyperparameters can affect the quality and diversity of the model's output, and can be useful for fine-tuning the model to specific tasks or applications.
The following hyperparameters are adjustable in GPT-3, GPT-3.5, and GPT-4:
→ Temperature [DEFAULT: 0.7]
Think of it as how "wild" or "tame" you want your AI friend to be when talking.
A higher temperature makes the AI more random and creative, while a lower one makes it more focused and sensible.
Example prompt: "What should I do on a sunny day?"
Temperature: 2.0 (high)
Hypothetical response: "Ride a giant ice cream cone to the moon and have a picnic with aliens!"
Explanation: The AI comes up with a very random and creative idea because of the high temperature.
Temperature: 0.5 (low)
Hypothetical response: "You can go outside and play with your friends, ride a bike, or have a picnic in the park."
Explanation: The AI gives a more focused and sensible answer because of the low temperature.
Accepted value: 0.0001 to infinity
→ Top K [DEFAULT: 40]
Imagine your AI friend has a list of words it can use to reply. Top K tells it how many words it can pick from the top of that list.
A smaller Top K makes the AI more focused, while a larger Top K lets it be more random.
Example prompt: "What is your favorite animal?"
Top K: 5 (low)
Hypothetical response: "My favorite animal is a dog."
Explanation: The AI chooses from a smaller set of words, so the answer is more focused and common.
Top K: 50 (high)
Hypothetical response: "My favorite animal is a pygmy marmoset."
Explanation: The AI chooses from a larger set of words, making the answer more diverse and less common.
Accepted value: 0 to infinity
→ Top P [DEFAULT: 0.9]
Top P is like a filter for your AI friend's words. It only picks words that have a combined chance of being picked that is less than or equal to the Top P value.
A higher Top P allows more variety, while a lower Top P makes the AI more focused.
Example prompt: "What's your favorite thing to do at the park?"
Top P: 0.95 (high)
Hypothetical response: "I love to feed the ducks, explore hidden paths, and look for unusual insects under rocks!"
Explanation: The AI has more variety in its answer because of the high Top P value.
Top P: 0.5 (low)
Hypothetical response: "I like to play on the swings."
Explanation: The AI gives a more focused and common answer due to the low Top P value.
Accepted value: 0 to 1
→ Repetition Penalty [DEFAULT: 1.0]
Imagine your AI friend talks a lot, but you don't want it to say the same thing over and over. Repetition Penalty helps you with that.
A higher penalty makes the AI avoid repeating words, while a lower penalty lets the AI repeat words more often.
Example prompt: "Tell me a story about a cat."
Repetition Penalty: 1.0 (low)
Hypothetical response: "There was a cat. The cat was fluffy. The cat liked to chase its tail. The cat was happy."
Explanation: The AI repeats the word "cat" often because of the low repetition penalty.
Repetition Penalty: 2.0 (high)
Hypothetical response: "Once upon a time, a fluffy feline lived in a cozy little house. This furry creature enjoyed chasing its tail and lived a joyful life."
Explanation: The AI avoids repeating the word "cat" and uses different words to describe it because of the high repetition penalty.
Accepted value: -infinity to infinity
Length Penalty [DEFAULT: 1.0]
Length Penalty is like a guide for your AI friend to make its answers longer or shorter.
A higher penalty makes the AI give shorter answers, while a lower penalty lets it give longer answers.
Example prompt: "What do you like about summer?"
Length Penalty: 2.0 (high)
Hypothetical response: "Sunshine."
Explanation: The AI gives a short answer because of the high length penalty.
Length Penalty: 0.5 (low)
Hypothetical response: "I love summer because of the warm sunshine, long days, outdoor activities like swimming and playing with friends, and delicious ice cream treats!"
Explanation: The AI gives a longer, more detailed answer due to the low length penalty.
Accepted value: 0 to infinity
→ Num Return Sequences [DEFAULT: 1]
Imagine you want your AI friend to give you different ideas about the same thing.
Num Return Sequences tells your AI friend how many different answers to give you.
Example prompt: "What can I do on a rainy day?"
Num Return Sequences: 1 (low)
Hypothetical response: "You can stay inside and read a good book."
Explanation: The AI gives only one answer because you asked for just one.
Num Return Sequences: 3 (high)
Hypothetical responses:
a) "You can stay inside and read a good book."
b) "You can watch a movie or play video games."
c) "You can try baking some cookies or making a craft."
Explanation: The AI gives three different answers because you asked for more ideas.
Accepted value: 1 to 10
→ Early Stopping [DEFAULT: FALSE]
Imagine your AI friend is telling you a story, but you don't want it to go on forever. Early Stopping is like a signal that tells your AI friend when to stop talking.
If you set an early stopping condition, the AI will stop once it reaches that point.
Example prompt: "Tell me a story about a dragon."
Early Stopping: True; Stop when the word "end" is generated.
Hypothetical response: "Once there was a dragon who lived in a cave. The dragon loved to fly in the sky and make friends with other animals. One day, the dragon met a unicorn, and they became best friends. The end."
Explanation: The AI stops the story once the word "end" is generated.
Note: Early Stopping is a stopping condition, not a parameter with high and low weights.
Accepted value: False // True; [insert specified condition]
→ No Repeat Ngram Size [DEFAULT: 0]
Imagine your AI friend sometimes says the same group of words again and again.
No Repeat Ngram Size tells your AI friend not to use the same group of words within a certain size (number of words) more than once.
Example prompt: "Tell me about your favorite sport."
No Repeat Ngram Size: 2 (low)
Hypothetical response: "I love playing soccer. Soccer is a fun sport that involves running, kicking the ball, and scoring goals. Soccer is played all around the world."
Explanation: The AI repeats the word "soccer" because the No Repeat Ngram Size is low (2), so it only avoids repeating groups of two words.
No Repeat Ngram Size: 4 (high)
Hypothetical response: "My favorite sport is soccer. It's a thrilling game that involves running, kicking the ball, and scoring goals. This activity is enjoyed by people all around the globe."
Explanation: The AI avoids repeating the word "soccer" in groups of four words, so it uses different words to describe the sport.
Accepted value: 0 to infinity
The following examples use differing prompts with varying hyperparameter conditions [copy/paste for your own research]:
→ EXAMPLE I
Prompt:
What are some good books to read for beginners?
Hyperparameters:
Temperature: 0.5
Top P: 0.5
Top K: 30
Repetition Penalty: 1.0
Length Penalty: 1.0
Number of Return Sequences: 1
Early Stopping: True; Stop when the phrase "Better start reading" is generated.
No Repeat N-gram Size: 4
→ EXAMPLE II
Prompt:
What are some easy recipes for dinner?
Hyperparameters:
Temperature: 0.7
Top P: 0.8
Top K: 50
Repetition Penalty: 1.5
Length Penalty: 1.2
Number of Return Sequences: 1
Early Stopping: True; Stop when the phrase "Enjoy your meal" is generated.
No Repeat N-gram Size: 3
→ EXAMPLE III
Prompt:
What are some effective ways to improve productivity?
Hyperparameters:
Temperature: 0.5
Top P: 0.9
Top K: 30
Repetition Penalty: 1.0
Length Penalty: 1.0
Number of Return Sequences: 1
Early Stopping: True; Stop when the phrase "Hope this helps" is generated.
No Repeat N-gram Size: 2
→ EXAMPLE IV
Prompt:
What are some fun activities to do on a weekend?
Hyperparameters:
Temperature: 0.9
Top P: 0.7
Top K: 20
Repetition Penalty: 1.2
Length Penalty: 1.3
Number of Return Sequences: 1
Early Stopping: True; Stop when the phrase "Have a great weekend" is generated.
No Repeat N-gram Size: 4
→ EXAMPLE V
Prompt:
What are some tips for staying motivated?
Hyperparameters:
Temperature: 0.5
Top P: 0.8
Top K: 40
Repetition Penalty: 1.0
Length Penalty: 1.0
Number of Return Sequences: 1
Early Stopping: True; Stop when the phrase "This is Sparta" is generated.
No Repeat N-gram Size: 2
Eye-Level Shot:
Sony Alpha a7 III camera with a Sony FE 24-105mm f/4 G OSS lens
Low-Angle Shot:
Sony Alpha a7 III camera with a Sony FE 16-35mm f/2.8 GM lens
High-Angle Shot:
Nikon D850 camera with a Nikkor 24-70mm f/2.8E ED VR lens
Extreme low-angle shot:
Canon EF 14mm f/2.8L II USM lens on a Canon EOS 5D Mark IV camera with a low-angle tripod
Extreme high-angle shot:
Canon EF 70-200mm f/2.8L IS III USM lens on a Canon EOS 5D Mark IV camera with a high-angle tripod
Side-Angle (Side-View):
Panasonic Lumix GH5S camera with a Panasonic Lumix 12-35mm f/2.8 II lens
Shot from Behind:
Nikon D850 camera with a Nikkor 70-200mm f/2.8E FL ED VR lens
Close-up shot:
Canon EF 50mm f/1.2L USM lens on a Canon EOS 5D Mark IV camera
Medium shot:
Canon EF 24-70mm f/2.8L II USM lens on a Canon EOS 5D Mark IV camera
Full shot:
Canon EF 16-35mm f/2.8L III USM lens on a Canon EOS 5D Mark IV camera
Extreme Close-Up Shot:
Canon EF 100mm f/2.8L Macro IS USM lens IS USM lens
Ground-Level Shot:
Canon EOS-1DX Mark III camera with a Canon EF 24-70mm f/2.8L II USM lens
Aerial Shot (Bird’s-Eye View):
DJI Phantom 4 Pro drone camera
Webcam-Style Shot:
Logitech C920 HD Pro Webcam
GoPro-Style:
GoPro HERO9 Black camera with GoPro Super Suit Dive Housing
Underwater Shot:
Canon EOS-1D X Mark II camera with a Canon EF 8-15mm f/4L Fisheye USM lens in an underwater housing
Underwater Close-Up Shot:
Canon EOS 5D Mark IV camera with a Canon EF 100mm f/2.8L Macro
Sunny: Canon EOS 5D Mark IV camera with a Canon EF 24-70mm f/2.8L II USM lens
Partly Cloudy: Nikon D850 camera with a Nikkor 24-70mm f/2.8E ED VR lens
Rainy: Sony Alpha a7 III camera with a Sony FE 16-35mm f/2.8 GM lens and waterproof housing
Snowy: Canon EOS-1D X Mark II camera with a Canon EF 24-105mm f/4L IS II USM lens and waterproof housing
Overcast: Panasonic Lumix GH5S camera with a Panasonic Lumix 12-35mm f/2.8 II lens
Foggy: Sony Alpha a7R IV camera with a Sony FE 70-200mm f/2.8 GM OSS lens
Hazy: Canon EOS R6 camera with a Canon RF 24-105mm f/4L IS USM lens
→TIME OF THE DAY ⏲️
Sunrise: Sony A7R IV camera with a Canon RF 15-35mm f/2.8L IS USM lens
Morning: Fujifilm X-T4 camera with a Fujinon XF 16-55mm f/2.8 R LM WR lens
Afternoon: Sony A7 III camera with a Sony FE 70-200mm f/2.8 GM OSS lens
Golden Hour: Canon EOS R5 camera with a Canon RF 50mm f/1.2L USM lens
Sunset: Sony A7R IV camera with a Sony FE 16-35mm f/2.8 GM lens
Twilight: Panasonic Lumix S1H camera with a Panasonic Lumix S Pro 50mm f/1.4 lens
Night: Sony A7S III camera with a Sigma 35mm f/1.2 DG DN Art lens
Dark and moody: Sony a7S III with Sony FE 50mm f/1.4 ZA lens
Bright and vibrant: Canon EOS R6 with Canon EF 16-35mm f/2.8L III USM lens
Fantasy: Fujifilm X-T4 with Fujinon XF 56mm f/1.2 R lens
Historical: Fujifilm GFX 100S with Fujinon GF 23mm f/4 R LM WR lens or Nikon Z7 II with Nikkor Z 50mm f/1.2 S lens
Adventure/Action: GoPro HERO10 Black
Documentary: Canon EOS C300 Mark III with Canon CN-E 24mm T1.5 L F lens
Horror: Blackmagic Pocket Cinema Camera 6K Pro with Sigma 18-35mm f/1.8 DC HSM Art lens
Mystical: Nikon D6 with Nikon AF-S NIKKOR 24mm f/1.4G ED lens
Romantic: Canon EOS R with Canon RF 50mm f/1.2L USM lens
Rustic: Panasonic Lumix GH5S with Sigma 18-35mm f/1.8 DC HSM Art lens
If you are looking to add some style and atmosphere to your images, you can add a filmmaker's name to your prompt.
If you start with a simple prompt like this one :
"A woman in front of her house in a scene from a movie",
you will get very varied and random results, without much artistic consistency.
On the other hand, if you want to add, for example, the name of Peter Jackson to the same prompt:
"A woman in front of her house in a scene from a movie directed by Peter Jackson",
the results will be completely different and your images will be tinged with the style of this filmmaker, through more coarse elements such as the scenery, the architecture, the clothes worn or sometimes more subtle ones such as the camera shots, the lighting, the colors and the general atmosphere.
You can be even more precise and add the title of one of the director's movies. In this example it could be "The Lord of the Rings".
I did some tests and I noticed that adding the name of the movie alone seems to give less refined results than adding the name of the director alone, which seems to give something more coherent and harmonious.
I invite you to test this and share your results to see if you notice the same or something different.
Finally, and this is even more subtle, but it seems that adding both the name of the director and the name of one of his movies in the same prompt adds a little more to the results.
You can take it a step further by adding the genre of the movie (for example Science Fiction), the date the movie was released and even the camera and lens used for a particular scene you want to reproduce.
That said, adding a filmmaker's name to your prompt won't magically transform your images into an amazing scene but it will definitely add a special touch to them and it's really fun to try.
I have compiled a list of some of the greatest filmmakers, sorted by film genre. This list is obviously not complete, but it is a good starting point!
→Action:
James Cameron (Terminator, Aliens)
Michael Bay (Bad Boys, Transformers)
John Woo (Hard Boiled, Mission: Impossible II)
→Comedy:
Woody Allen (Annie Hall, Manhattan)
Mel Brooks (Blazing Saddles, Young Frankenstein)
Charlie Chaplin (The Gold Rush, City Lights)
→Drama:
Martin Scorsese (Goodfellas, The Departed)
Francis Ford Coppola (The Godfather)
Akira Kurosawa (Seven Samurai, Rashomon)
→Fantasy:
Peter Jackson (The Lord of the Rings, The Hobbit)
George Lucas (Star Wars, Indiana Jones)
Guillermo del Toro (Pan's Labyrinth)
→Superhero:
Jon Favreau (Iron Man, The Mandalorian)
Sam Raimi (Spider-Man, Doctor Strange)
Patty Jenkins (Wonder Woman)
→Horror:
Alfred Hitchcock (Psycho, The Birds)
John Carpenter (Halloween, The Thing)
Wes Craven (A Nightmare on Elm Street, Scream)
George A. Romero (Night of the Living Dead)
→Musical:
Bob Fosse (Cabaret, All That Jazz)
Baz Luhrmann (Moulin Rouge, The Great Gatsby)
Gene Kelly (Singin' in the Rain, An American in Paris)
→Sci-Fi:
Ridley Scott (Alien, Blade Runner)
Stanley Kubrick (2001: A Space Odyssey)
Christopher Nolan (Inception, Interstellar)
Denis Villeneuve (Arrival, Dune)
James Cameron (Titanic, Avatar)
Steven Spielberg (E.T., Ready Player One)
→Thriller:
David Fincher (Se7en, Gone Girl)
Quentin Tarantino (Pulp Fiction, Kill Bill)
M. Night Shyamalan (The Sixth Sense, Split)
→Western:
Sergio Leone (The Good, the Bad and the Ugly)
John Ford (The Searchers, Stagecoach)
Clint Eastwood (Unforgiven, The Outlaw Josey Wales)
→Animation:
Hayao Miyazaki (Princess Mononoke)
Walt Disney (The Lion King)
*You can also add Pixar to your prompt. The results are often really great!
→Crime:
Brian De Palma (Scarface, The Untouchables)
Michael Mann (Heat, Collateral)
Quentin Tarantino (Pulp Fiction)
→Historical:
Steven Spielberg (Schindler's List)
James Cameron (Titanic)
Mel Gibson (Braveheart)