Summary: The video provides a comparative analysis of various AI image editors, including Nano Banana, ChatGPT, Quen ImageEdit, Flux Context, Reeve, and Cadream 4. The presenter tested each tool four times for various challenging use cases, focusing on their ability to remove objects, change scenes, maintain character consistency, and understand complex prompts. Cadream 4 and Reeve were highlighted for their realistic character generation and unique editing features, respectively. In contrast, Nano Banana was praised for its overall aesthetic and prompt understanding, making it a strong all-around choice despite not always being the best in specific tasks. The video also promoted HubSpot's AI-powered viral toolkit and Futuredia's AI courses.
AI Image Editor Comparison #
Removing Objects and Changing Scenes #
- Removing People (Difficult Image with Glass and Shadows):
- All successfully removed people but struggled with filling details.
- Chat GPT & Flux Context: Removed pillar completely.
- Seedream & Reeve: Removed people but didn't fill the beam.
- Nano Banana & Quen: Understood the pillar and filled it (different ways).
- Quen: Best with shadows, but also removed boats (unprompted). Winner: Quen (despite removing boats).
- Changing Daytime to Nighttime / Adding Milky Way:
- Cadream: Removed too much background.
- Nano Banana (Preference): Maintained details, turned off headlights for daytime.
- Chat GPT: Limited aspect ratios, often crops.
- Reeve & Context: Terrible.
- Seedream & GPT: Alright.
- Nano Banana & Quen: Amazing for night/Milky Way.
- Nano Banana (Winner): Natural shadows, unlike Quen.
Complex Prompt Understanding & Specific Details #
- Optimus Prime with Blender Model on Screen:
- All did surprisingly well retaining details.
- Nano Banana: Most natural/realistic, but jumbled text.
- Seedream (Preferred): Jumbled text, but looks like another language.
- Quen & JachiBT: Strong contenders.
- Aang in Avatar State with Elements:
- Most didn't perform as well.
- Nano Banana: Preferred style, but jumbled text (Nickelodeon) and missing floating elements.
- Chat GPT: More accurate overall, less preferred style.
- Swapping Clothes:
- Seedream & Reev: Only ones to correctly switch clothes.
- Nano Banana: Returned original image.
- Reeve: Removed background and changed poses/size.
- Seedream (Winner): Kept size, unzipped hoodie, kept a natural look.
- Restoring and Colorizing Damaged Photo:
- Flux Context: Failed.
- Chat GPT: Good people but characters don't resemble original.
- Seedream (Winner): Most accurate, especially close-up on faces.
- Nano Banana & Quen: Not bad.
- Drake Meme (Replacing Face & Text):
- Quen: Didn't try to replace face.
- Nano Banana: Small, unblended face.
- Reeve: Natural lighting, but altered face/poses.
- Chat GPT (Clear Winner): Closest face, better overall.
- Combining Multiple Items into Detailed Scene:
- Prompt required understanding of 8 uploaded items, positions, and details.
- All did a decent job, none completely failed.
- Context: Off-details, wrong hat, changed room.
- ChatGpt: Accurate to uploads (except faces slightly off).
- Quen: Many details off, unnatural.
- Nano Banana: Characters too small, missing sunglasses, wrong jacket, poor faces.
- Seedream (Clear Winner): Closest details, natural lighting, fit scene best, best faces.
- Extracting Outfits:
- Nano Banana: Left hands.
- Seedream: Didn't try.
- Quen: Added yellow stripe.
- Chat GPT: Forgot sleeves.
- Flux Context: Left off boots.
- Reeve (Only one that got it right): Accurate outfit, retained pose.
- Swapping Firebender/Airbender Clothes (from extracted outfit):
- All understood the goal.
- Reeve (Winner): Included all outfit parts, retained pose.
- Adding Monster Outside Window (from video project):
- Quen: Failed.
- Other four: Did a good job.
- Seedream: Monsters Inc. inspiration. No clear winner, all were decent.
- Rhino with Red Glow on One Horn (Specific detail):
- Nano Banana & Quen: Couldn't apply glow to one horn.
- Seedream: Wrong horn.
- Chat GPT: Right horn, but changed style too much.
- Flux Context: Unclear generation.
- Reeve (Nailed it): Correct angle and horn highlight.
- "Show What Red Arrow Sees" (Google Maps):
- Quen, Seedream, Flux: Failed.
- Chat GPT & Nano Banana: Accurate depending on angle.
- Reeve: Decent, less accurate to building.
- "Photo from Ground View, More Zoomed In" (Google Maps):
- Quen: Wouldn't generate.
- Same three failed, same three attempted well.
- Nano Banana (Winner): Best attempt.
- "Make France Red with Glowing Outline, Label as France" (Map):
- All failed.
- Nano Banana, Quen, Reeve: Decent shape of France.
- Nano Banana: Best shape, but placed on top of everything.
- "Add Eiffel Tower Popping Up on Map in 3D (Correct Location)":
- None placed it in the right spot.
- Seedream: Changed style completely.
- Reeve (Winner with Editing Feature): Unique feature to move segmented Eiffel Tower to correct spot.
Consistent Characters (Human, Celebrity, Animation Styles, Product) #
- Man Surfing:
- Flux Context: Good, but low quality face resemblance.
- Chat GPT: Attempted resemblance, but created new character, changed details, 4 toes.
- Quen: Resembles, but not much like the person.
- Nano Banana: Good job on face (though lower resolution), best overall image/blend.
- Seedream: Crystal clear details, exact likeness, but base image model less good (weird wave, not perfectly blended).
- Reeve: Exact likeness, good shot (GoPro like).
- Usable for likeness (Winner): Seedream and Reeve.
- Man Skydiving/Volcano Boarding/Hang Gliding:
- Quen & Chat GPT: Poor resemblance consistently.
- Nano Banana: Decent likeness (hard to tell with helmet), good image.
- Seedream: Perfect likeness, good aesthetics.
- Reeve: Nailed likeness (teeth sometimes weird).
- Flux Context: Good likeness, low resolution, sometimes weird shot composition.
- Consistent Character Summary (Real Person):
- Seedream (Best): Absolute best resemblance.
- Reeve: Close second, but hit or miss on composition.
- Flux Context: Great consistency, lower image quality.
- Nano Banana: Medium consistency, best base image.
- Chat GPT: Never looks like person, changes details, okay (not realistic) images.
- Quen: Usually terrible.
- Man in Winter Scene (with reference photo):
- Flux Context: Warped face.
- Chat GPT: Terrible.
- Quen: Not bad, but not detailed.
- Nano Banana: Better detail.
- Seedream: Terrible perspective, but good image from photo.
- Reeve (Winner): Especially helped by reference, retained likeness, natural pose, aesthetic.
- Nano Banana (for exact retention): Kept original photo details.
- Celebrity (Danny Devito):
- Nano Banana & Seedream: Close likeness, Nano Banana more natural.
- Quen & ChachiBT: Terrible.
- Reeve: Alright likeness, weird lighting.
- Context: Decent likeness, natural shot (with Nano Banana).
- Minecraft Character:
- Flux Context: Didn't do well.
- All others: Good job.
- Seedream: Too detailed for Minecraft.
- Chachib (Best):
- South Park Character:
- Flux Context: Wouldn't do it.
- Nano Banana: Looks like person.
- Chachib (Best for South Park vibe):
- Claymation Character:
- Quen & Flux Context: Kind of rude, doesn't look much like person.
- Nano Banana, Seedream, Chachi BT: Great job.
- Reeve: Funny style.
- Action Figure in Box:
- All okay.
- Nano Banana: Best.
- GTA 5 Character:
- Nano Banana (Winner):
- ChachiBT & Reeve: Close.
- Quinn & Cadream: Not bad.
- Flux Context: Wouldn't generate.
- Rubber Hose Animated Character:
- Most didn't understand style.
- JBT & Reev: Understood, but not great looks.
- Nano Banana (Perfect Representation, Winner):
- Movie Poster (Parody: The Good, The Bad, and The AI):
- All did okay.
- Quinn: Not enough likeness.
- Reeve: Weird, copy-pasted face.
- Context: Solid, good text.
- Chai BT & Seedream: Great posters, Seedream looked more like person.
- Nano Banana (Amazing, Far Best Design): Unique font, solid text, perfect theme.
- Grid of Man with Different Emotions (3x2: Happy, Sad, Excited, Scared, Embarrassed, Angry):
- Reeve: Didn't understand, made face weird.
- Lux Context: Alright, correct emotions, didn't retain likeness.
- Tragic: No likeness.
- Quen: Not much likeness.
- Nano Banana: Good job on face (not perfect), correct emotions.
- Cadream (Winner): Most likeness, exaggerated emotions.
- Consistent Woman / Riding Iceberg / Spaceship on Mars:
- Similar results: Chat GPT & Quen not good, Nano Banana good likeness & best overall image, Cadream highest likeness but harder to follow prompt, Reeve & Context also good.
- Woman & Man Having Drink at Bar:
- Gwen & ChachiT: Bad.
- Context: Decent, but small details off.
- Reeve: Lost likeness/quality, extra nose ring.
- Seedream: Perfect likeness.
- Nano Banana: Really good aesthetic and likeness, good vibe.
- Converting MidJourney Image to Different Styles:
- 3D Animation: Seedream nailed it, Chacht alright.
- Anime: Everything but Nano Banana okay, Context & Quen best.
- Street Art Graffiti: All but Seed Dream did good (different interpretations). Nano Banana was exceptionally talented.
- Zooming Out (Full Body Shot):
- Chat GBT: Complete fail.
- Seedream: Didn't zoom out enough.
- Reeve & Quen: Weird images, not standing on anything.
- Nano Banana & Flux Context (Winner): Best retention of face, colors. Nano Banana better outfit.
- Zooming In (Close-up):
- Nano Banana (Best): Basically a cropped version.
- Seedream, Reeve, Flux Context: Also good close-ups. Tied.
- Changing Camera Angle (Mongolian Horseman Front View Full Body):
- Quinn & Nano Banana: Got full body shot right.
- Sedream: Amazing character consistency but less adherence to prompt.
- Character Turnaround Sheet:
- JGBT: Fail.
- Context, Seedream, Nano Banana, Quen: All did awesome, kept details consistent from different angles. Three/Four-way tie.
- Reeve: Completely changed style.
- Character Turnaround Sheet (Retain Original Pose):
- Reeve (Winner): Only one to retain pose, nailed character.
- Context, Quen, Nano Banana: Good character likeness.
- Seedream & Chat GPT: Changed style too much.
- Consistent Product (Replacing Can):
- Nano Banana, Quen, Cdream, Chat GPT: Missing details, wrong details, or changed image. Flux failed.
- Reeve (Closest): Got logo, but not all text.
- Consistent Product (Larger Can):
- Nano Banana: Pretty close.
- See: Huge can to get details in.
- Quen: Looks good from far, not close.
- Chat GPT: Changed details, but got "future fuel."
- Reeve (Winner): Got "future fuel" and logo, natural can size.
- Cadream: All details, but huge can.
- Consistent Product (Again):
- Nano Banana, Seedream, Reeve, Flux Context: Got all details.
- Quen: Bad text, wrong can size.
- Nano Banana & Reeve (Tied for Aesthetics): Passable. Reeve wins for no black bars.
- Modern Landing Page for Product:
- Nano Banana (Winner): Almost all text correct, great looking site, looks like actual landing page.
- Cream: Not a landing page, just huge design.
- Quen: Ambitious text, failed.
- ChachiBT: Simple, all text accurate (knew its limits).
- Reeve: Weird.
- Context: Lots of bad text.
- Change Landing Page Text to Chinese:
- Reeve: Converted all text, but translations were weird/random.
- Quen: Didn't translate much, but accurate.
- Context: Only part real characters.
- Cadream & ChatgPT (Winners): Normal translation, real translations on what they did translate.
- Nano Banana: Not bad, but little off.
Specific Artistic Styles & Pose Adherence #
- Stick Figure Fighting Pose to Fight Scene (Skeleton Nallerina vs Pimp Cat):
- Quen (Definitely the Best): Got poses basically perfect, consistent across environments (disco, casino, alley).
- Nano Banana: Tried (for pose), winner for aesthetics and character consistency in great scenes (actually fighting).
- Other four: Didn't do good for pose or scene aesthetics.
Consistent Scene & Character Movement (Anime Market) #
- Higher Angle of Scene:
- Nano Banana: Same original image.
- Cedream & Flux Context: Good. Between these two.
- Full Bird's Eye View/Top-Down Aerial:
- Cadream: Failed.
- JBT: Completely wrong style.
- Nano Banana (Best): Most original shot feel.
- Quen & Context: Not bad. Reeve on right track but wrong style.
- Close-up of Woman (Straight-on shot):
- Flux Context & Nano Banana: Would work.
- Others: Changed character or shot.
- Zooming in on Man:
- Flux Context & Nano Banana: Great job.
- Cadream: Looks good, but doesn't understand scene.
- Reeve & Chad: Changed character.
- Gwen: Okay, but didn't fully follow prompt.
- Zooming in on Cat:
- Flux Context: Really weird.
- Most others would work (preference based on film shot).
- Grid of Woman/Man/Cat with Different Emotions:
- Nano Banana, Seedream, Flux Context: Good consistent characters.
- Nano Banana (Very Clear Winner for Emotions): Nailed all emotions perfectly.
Tool by Tool Summary #
- Quen:
- Okay at many challenges.
- Winner: Stick figure pose adherence.
- Tied for removing people.
- Passable in many, rarely winner.
- Reeve:
- Most surprising.
- Good at "thinking things through" challenges (Rhino).
- Good for consistent character (Drake face swap, character sheet pose).
- Favorite Feature: Platform editing (moving objects).
- Flux Context:
- Older model, few wins.
- Best at animation styles with consistent characters and scenes (moving camera/changing environment).
- Chat GPT:
- Probably best at prompt understanding.
- Often changes character details.
- Disliked style, limited aspect ratios.
- Not realistic.
- Cadream 4 (Seedream):
- Not strong on prompt adherence.
- Good for realistic consistent characters where fine details matter.
- Benefit: Generates in 4K.
- Nano Banana:
- Overall Recommendation if picking one tool.
- Good aesthetics, good prompt understanding, good consistent characters, good graphic design.
- Weakness: Not always the best for specific use cases.
The video also mentions a free AI-powered viral toolkit from HubSpot and Futuredia's AI courses, including a full course on Nano Banana.
last updated: