Frequently asked questions
AI Virtual Try-On in Instagram DMs
A user comments on a post or clicks an Instagram ad, then receives a DM from
your account.
There are two main entry points:
1. Comment-to-DM (Organic posts & Reels)
User leaves a comment under your post.
This comment triggers a DM from your account.
2. Click-to-DM Instagram Ads (Keyword-based)
A user clicks your Instagram ad and sees a pre-filled message with a keyword (for example, Dress_12K).
When they tap Send, the keyword is sent to your DMs and triggers a specific try-on flow.
Inside the DM, the user uploads a photo, and AI generates a try-on image using the user's real garment image and the user’s photo.
Everything happens inside Instagram DMs.
There are two main entry points:
1. Comment-to-DM (Organic posts & Reels)
User leaves a comment under your post.
This comment triggers a DM from your account.
2. Click-to-DM Instagram Ads (Keyword-based)
A user clicks your Instagram ad and sees a pre-filled message with a keyword (for example, Dress_12K).
When they tap Send, the keyword is sent to your DMs and triggers a specific try-on flow.
Inside the DM, the user uploads a photo, and AI generates a try-on image using the user's real garment image and the user’s photo.
Everything happens inside Instagram DMs.
AI uses your real garment image.
The quality of the result depends heavily on how clear and understandable your garment photo is.
The quality of the result depends heavily on how clear and understandable your garment photo is.
Yes. AI keeps the user’s real face. Small visual changes can happen in rare
cases, but the face is preserved.
AI does not understand real body measurements.
It works visually, similar to Photoshop, not 3D modeling.
Small differences can appear.
It works visually, similar to Photoshop, not 3D modeling.
Small differences can appear.
No. AI does not understand sizes.
It visually places the garment based on the photo.
It visually places the garment based on the photo.
Yes. Users do not need to install another app or open a website.
Everything happens inside Instagram DMs.
Everything happens inside Instagram DMs.
No. Setup takes only a few minutes and requires no coding skills.
Yes. Most brands test by commenting on their own post from a different account
(for example, a personal account).
Apiway also provides free test try-ons.
Apiway also provides free test try-ons.
Yes. Apiway is approved by Meta and works through the official Meta API.
You can disconnect it anytime in Instagram → Apps and Websites.
You can disconnect it anytime in Instagram → Apps and Websites.
It works as both:
People spend more time with the product, share try-on results, and move into DMs — which increases buying intent.
This is not just an experiment anymore.
Zara has already added AI Virtual Try-On inside their app, which clearly shows this is a real trend.
AI Try-On is becoming a new standard in fashion, and most brands will adopt it soon. Brands that start earlier get an innovation advantage.
- a strong engagement tool
- a conversion booster
People spend more time with the product, share try-on results, and move into DMs — which increases buying intent.
This is not just an experiment anymore.
Zara has already added AI Virtual Try-On inside their app, which clearly shows this is a real trend.
AI Try-On is becoming a new standard in fashion, and most brands will adopt it soon. Brands that start earlier get an innovation advantage.
Set Up AI Virtual Try-On on Apiway
You need an Instagram Business account or a Creator account.
Personal accounts are not supported.
Personal accounts are not supported.
You connect your Instagram account by logging in through Apiway and granting
access via the official Meta connection flow.
This takes less than a minute.
This takes less than a minute.
No. You only need:
For organic posts and comment triggers, Ads Manager is not required.
- a Business or Creator Instagram account
For organic posts and comment triggers, Ads Manager is not required.
Yes. An Instagram Shop and product tags are not required.
AI Virtual Try-On works independently from Instagram Shopping.
AI Virtual Try-On works independently from Instagram Shopping.
Yes.
You can link try-on flows to any website, including:
You can link try-on flows to any website, including:
- Shopify
- WooCommerce
- custom websites
- landing pages
Inside the try-on flow settings, there is a field where you can paste:
This link appears as a button inside the DM flow, right after the try-on result.
- a product page link
- a checkout link
- or another Instagram profile
This link appears as a button inside the DM flow, right after the try-on result.
You can manage basic actions from your phone, but for setup and customization it
is better to use a desktop.
A desktop is faster and more comfortable for creating flows, editing prompts, and managing settings.
A desktop is faster and more comfortable for creating flows, editing prompts, and managing settings.
Apiway connects directly to Instagram through the Meta API.
You do not need tools like ManyChat.
Apiway is a ManyChat alternative, built specifically for the fashion industry.
You do not need tools like ManyChat.
Apiway is a ManyChat alternative, built specifically for the fashion industry.
No.
Currently, one Apiway account can be connected to one Instagram account only.
Currently, one Apiway account can be connected to one Instagram account only.
There is no hard limit.
You can create as many try-on flows as you need for posts, ads, and campaigns.
You can create as many try-on flows as you need for posts, ads, and campaigns.
Keywords, Flows, Ads vs Comments
Click-to-DM Ads
Comment-to-DM Triggers
- Triggered from Instagram Ads
- User clicks the ad and sends a keyword in DMs
- Each keyword triggers a specific flow
- Best for paid traffic and precise control
Comment-to-DM Triggers
- Triggered when a user comments under a post or Reel
- The keyword is just a trigger
- The post itself defines which flow runs
- Best for organic content and influencer collaborations
Garment / Product Images
The most important thing is that the garment is clear and visually easy to
understand.
Best results come from:
The clearer the garment shape, the better the AI result.
Best results come from:
- front-facing angle
- neutral pose
- even lighting
- no extreme shadows
- no motion blur
The clearer the garment shape, the better the AI result.
All three can work, but results differ:
If you want maximum consistency, a ghost mannequin is recommended.
- Ghost mannequin → best and most stable results
- Real model (neutral pose) → good results
- Flat lay → works, but less accurate for fit
If you want maximum consistency, a ghost mannequin is recommended.
Yes.
Most brands use the same images they already use for:
Just avoid very artistic or extreme poses.
Most brands use the same images they already use for:
- Product pages
- Instagram posts
- Ads
Just avoid very artistic or extreme poses.
Both work.
If the garment shape is clear, the background is not critical.
- White or neutral background → more predictable results
- Lifestyle background → works, but AI may guess missing details
If the garment shape is clear, the background is not critical.
You do not need to remove the background.
AI handles the garment extraction automatically.
AI handles the garment extraction automatically.
Supported formats:
WebP is not recommended.
- JPG / JPEG
- PNG
WebP is not recommended.
Recommended:
Higher-quality images always give better results.
- minimum 1024px on the longest side
- sharp, non-compressed image
Higher-quality images always give better results.
Yes, but results may vary.
Photorealistic 3D renders work better than stylized ones.
Real photos usually give more predictable results.
Photorealistic 3D renders work better than stylized ones.
Real photos usually give more predictable results.
These fabrics are challenging.
AI may:
Using a ghost mannequin and clear instructions like “reduce transparency” improves results.
AI may:
- reduce transparency
- simplify details
- change how the fabric looks
Using a ghost mannequin and clear instructions like “reduce transparency” improves results.
AI can handle shiny fabrics, but:
Good lighting in the original image helps a lot.
- Reflections may look different
- Highlights can be simplified
Good lighting in the original image helps a lot.
Yes, in most cases.
Small logos and complex patterns may slightly change, especially if the source image is low quality.
Small logos and complex patterns may slightly change, especially if the source image is low quality.
Long and tight dresses are one of the hardest cases.
Results depend heavily on:
Newer models handle this better, but occasional issues can happen.
Results depend heavily on:
- garment image clarity
- user photo pose
Newer models handle this better, but occasional issues can happen.
Not always.
Sometimes AI:
This happens more often with older models and improves with newer ones.
Sometimes AI:
- Removes previous clothing correctly
- Sometimes layer garments incorrectly
This happens more often with older models and improves with newer ones.
Because users upload different photos.
AI adapts the garment to:
If the input photo is unclear, AI may guess missing details.
AI adapts the garment to:
- pose
- lighting
- body angle
If the input photo is unclear, AI may guess missing details.
Minimum:
Optional (for better results):
- 1 clear front image
Optional (for better results):
- back view
- side view
Yes.
Each color should be treated as a separate garment with its own image.
Each color should be treated as a separate garment with its own image.
Yes, but this is an advanced setup.
Best practice:
Best practice:
- Start with one main garment
- Then experiment with full sets
User Photos & Experience in DMs
Best results come from:
The clearer the photo, the better the result.
- full-body or half-body photos
- clear lighting
- front-facing pose
- no heavy filters
The clearer the photo, the better the result.
Any photo can work, but mirror selfies work best.
Most users naturally send mirror selfies, even if you don’t ask for them.
That’s why most try-on flows are designed around mirror selfies.
Most users naturally send mirror selfies, even if you don’t ask for them.
That’s why most try-on flows are designed around mirror selfies.
No.
A plain background helps, but it is not required.
AI focuses on the person, not the background.
A plain background helps, but it is not required.
AI focuses on the person, not the background.
AI will still try to generate a result, but quality will drop.
Poor lighting and blur increase the chance of:
Poor lighting and blur increase the chance of:
- Wrong garment shape
- Missing details
- Rejected images
This can happen due to:
Sometimes re-uploading the same photo works.
- very low image quality
- heavy filters
- extreme angles
- content safety filters
Sometimes re-uploading the same photo works.
AI has protection systems to avoid using celebrity images.
If a photo looks very professional or similar to public figures, it may be blocked by mistake.
If a photo looks very professional or similar to public figures, it may be blocked by mistake.
Yes, it can work.
Standing photos give better results, but sitting photos are supported.
Standing photos give better results, but sitting photos are supported.
It can work, but the results are less predictable.
Front-facing photos give the best results.
Front-facing photos give the best results.
No. Only one person per photo is supported.
AI may reject the image.
Photos with nudity, underwear, or high transparency can trigger safety filters.
Photos with nudity, underwear, or high transparency can trigger safety filters.
Yes. AI keeps the user’s real skin tone.
Small lighting variations can happen depending on the photo.
Small lighting variations can happen depending on the photo.
AI will still generate an image.
Results may look unusual, but the system does not block based on gender or category.
Results may look unusual, but the system does not block based on gender or category.
How do I explain to users how to take a good photo?
Best approach:
Short instructions work better than long explanations.
Best approach:
- Add a short instruction inside the DM
- Show a simple example image
Short instructions work better than long explanations.
They see a loading state inside the DM while the image is being generated.
Usually 10–30 seconds, depending on model load and image complexity.
Yes. Users can upload multiple photos and retry as many times as they want.
Yes. You can design flows that allow users to try different items in the same DM
thread.
Yes. Users can save images directly from Instagram like any other image.
Yes. Users can share try-on images to Stories, Feed, or send them to friends.
You can add a small watermark or branding to the generated image.
Yes. You can ask users for:
This can be added as a step in the DM flow.
- Phone number
- Other contact info
This can be added as a step in the DM flow.
Yes. You can add a button with a product or checkout link directly after the
try-on result.
AI Behaviour, Quality & “Hallucinations”
AI mistakes are part of today’s reality.
From real usage:
In most cases, retrying the same photo or uploading a new selfie fixes the issue.
From real usage:
- Nano Banana 1: noticeable issues happen roughly 1 out of 40–50 try-ons
- Nano Banana 2: issues are much rarer, about 1 out of 80–100 try-ons
In most cases, retrying the same photo or uploading a new selfie fixes the issue.
AI does not understand real body anatomy or clothing sizes.
It works like advanced image editing, not real 3D modelling.
Because of this, AI may slightly change:
Usually, these changes stay within 5–10%, but they can be noticeable.
It works like advanced image editing, not real 3D modelling.
Because of this, AI may slightly change:
- Body proportions
- Leg width
- Waist or shoulders
Usually, these changes stay within 5–10%, but they can be noticeable.
AI can confuse faces.
This happens because of:
This is a technical limitation, not a setup issue.
Nano Banana 1 shows this more often.
Nano Banana 2 reduces it significantly, but AI is still not perfect.
This happens because of:
- Lighting and shadows
- Unusual angles
- Heavy filters
- AI safety systems are trying to block celebrity misuse
This is a technical limitation, not a setup issue.
Nano Banana 1 shows this more often.
Nano Banana 2 reduces it significantly, but AI is still not perfect.
In most cases, the hairstyle is preserved.
However, hair can sometimes:
Long hair and complex poses increase the chance of glitches.
However, hair can sometimes:
- overlap straps or collars
- cover parts of the garment
- clip through clothing
Long hair and complex poses increase the chance of glitches.
AI struggles with clothing logic and layering.
It does not always understand:
For example:
Nano Banana 1 struggles with this more often.
Nano Banana 2 handles this much better, but mistakes can still happen.
It does not always understand:
- When clothes should replace existing garments
- When they should be layered on top
For example:
- Dresses should replace pants
- Jackets should go over T-shirts
Nano Banana 1 struggles with this more often.
Nano Banana 2 handles this much better, but mistakes can still happen.
Prompts can reduce mistakes, but they cannot fully control AI behavior.
AI may still:
Prompt tuning helps, but it is not a 100% guarantee.
AI may still:
- Ignore instructions
- Interpret them incorrectly
Prompt tuning helps, but it is not a 100% guarantee.
Usually yes, but not always.
AI tries to adapt garment lighting to the selfie.
If the user's photo has poor lighting, the garment will also look worse.
Better user photos = better results.
AI tries to adapt garment lighting to the selfie.
If the user's photo has poor lighting, the garment will also look worse.
Better user photos = better results.
Small color shifts can happen due to:
White, black, and metallic garments are the most sensitive.
Clear product photos reduce this issue.
- Lighting differences
- Shadows
- Reflective or shiny fabrics
White, black, and metallic garments are the most sensitive.
Clear product photos reduce this issue.
This happens when AI lacks visual information and starts guessing.
Common reasons:
Retrying often fixes the problem.
Nano Banana 2 significantly reduces these artifacts.
Common reasons:
- low-quality garment photos
- complex fabrics
- poor user selfies
Retrying often fixes the problem.
Nano Banana 2 significantly reduces these artifacts.
Lower-body garments are the weakest category.
Pants, skirts, and shorts fail more often because:
We recommend avoiding lower-body-only try-ons.
Pants, skirts, and shorts fail more often because:
- Legs are often covered
- Poses vary too much
- AI struggles to remove existing clothing
We recommend avoiding lower-body-only try-ons.
This is one of the hardest scenarios.
If a user:
AI must “bring legs together,” which Nano Banana 1 often fails at.
Nano Banana 2 works better, but errors can still occur.
If a user:
- wears jeans
- stands with legs apart
AI must “bring legs together,” which Nano Banana 1 often fails at.
Nano Banana 2 works better, but errors can still occur.
AI works, but results are less predictable.
Reasons:
Results are improving, but consistency is lower than with standard adult body types.
Reasons:
- Fewer training examples
- Greater body variation
Results are improving, but consistency is lower than with standard adult body types.
No.
AI does not understand clothing sizes.
It does not know what S, M, or XL means.
AI only works visually, not logically.
AI does not understand clothing sizes.
It does not know what S, M, or XL means.
AI only works visually, not logically.
No.
AI generates one image per photo.
It cannot reliably generate accurate side or back views from a single selfie.
AI generates one image per photo.
It cannot reliably generate accurate side or back views from a single selfie.
Failures are rare but expected.
In most cases:
This is normal behaviour for AI systems today.
In most cases:
- Retrying works
- Changing the selfie works even better
This is normal behaviour for AI systems today.
Yes — but not perfect.
Best practices:
Even then, AI will never be 100% perfect today.
Best practices:
- Use clear garment photos
- Use a ghost mannequin when possible
- Guide users to good selfies
- Use Nano Banana 2
Even then, AI will never be 100% perfect today.
Models, Prompts & Customization
The main difference is stability, accuracy, and error rate.
Nano Banana 1.0
Nano Banana 2.0
Both models work, but Nano Banana 2.0 clearly produces cleaner and more consistent results.
Nano Banana 1.0
- Cheaper
- Faster
- More hallucinations
- More problems with layering (shirt over pants, hair glitches, tight dresses)
- Good for simple garments and testing
Nano Banana 2.0
- More expensive
- Much more stable
- Better face preservation
- Better garment accuracy
- Better handling of complex cases (tight dresses, poses, long garments)
Both models work, but Nano Banana 2.0 clearly produces cleaner and more consistent results.
If you are:
If you:
Many brands start with 1.0 and upgrade later.
- testing the feature
- working with simple garments (T-shirts, jackets, bags)
- running a small campaign
If you:
- sell dresses or complex garments
- care a lot about visual quality
- plan to scale try-ons seriously
Many brands start with 1.0 and upgrade later.
Because it:
In practice, you spend less time fixing problems, so the higher cost often pays off.
- uses a more advanced model
- makes fewer mistakes
- needs more computing power
- produces fewer failed generations
In practice, you spend less time fixing problems, so the higher cost often pays off.
Inside each Try-On Flow in Apiway, there is a Prompt section.
You can:
No coding required.
You can:
- Use the default prompt
- Edit it
- Replace it fully
- Duplicate flows with different prompts
No coding required.
Yes.
Apiway includes prebuilt prompts optimized for:
These prompts are designed for the Overlay strategy and work well for most brands.
You can customize them if needed, but the default prompts already work.
Apiway includes prebuilt prompts optimized for:
- Tops
- Dresses
- Outerwear
- Accessories
These prompts are designed for the Overlay strategy and work well for most brands.
You can customize them if needed, but the default prompts already work.
Overlay Strategy
New Image Strategy
- Uses the user’s real photo
- Replaces only the garment area
- Keeps face and background
- More realistic for users
- More sensitive to bad selfies
New Image Strategy
- Generates a new image based on the user’s face
- Can change background and lighting
- The garment often looks “perfect”
- Face and body can look less like the real person
From real usage:
- Women’s clothing → Overlay strategy works better
- Men’s clothing → New Image strategy often works better
You can add instructions like:
This helps AI avoid mistakes, but it may slightly change the garment's appearance.
This is a trade-off between accuracy and stability.
- “Exclude transparency”
- “No sheer fabrics”
- “Solid, opaque fabric only”
This helps AI avoid mistakes, but it may slightly change the garment's appearance.
This is a trade-off between accuracy and stability.
You can try, but it’s not guaranteed.
AI still struggles with clothing removal logic:
AI still struggles with clothing removal logic:
- Sometimes it removes pants
- Sometimes it layers over them
Yes — with New Image strategy.
You can describe:
The overlay strategy keeps the original background.
You can describe:
- City
- Studio
- Lifestyle scene
- Creative environments
The overlay strategy keeps the original background.
Yes.
That’s exactly how the Overlay strategy works:
That’s exactly how the Overlay strategy works:
- The background stays untouched
- Only the garment area is modified
Text overlays are not generated by AI itself.
If needed, you should:
Apiway focuses on image generation, not post-design.
If needed, you should:
- Add text after generation
- Or use external tools for branding
Apiway focuses on image generation, not post-design.
Yes.
You can:
You can:
- Include watermark instructions in the prompt
- or add branding after generation
Yes, but they are generous.
Very long or overly complex prompts can:
Very long or overly complex prompts can:
- slow generation
- increase error rate
- confuse the model
Yes.
You can:
You can:
- Duplicate flows
- Use different prompts
- Compare results manually
Not yet in a true “training” sense.
AI does not learn from your brand automatically.
However, you can:
AI does not learn from your brand automatically.
However, you can:
- Refine prompts
- Standardize garment photos
- Reuse working templates
Performance, Limits & Reliability
Apiway is built on stable cloud infrastructure and is designed to run 24/7.
Like any system that depends on external platforms, short interruptions can happen due to:
In real usage, uptime is high, and most issues are temporary.
Like any system that depends on external platforms, short interruptions can happen due to:
- Instagram (Meta) API issues
- AI model provider availability
- Global traffic spikes
In real usage, uptime is high, and most issues are temporary.
Nothing breaks.
Each user:
The system is designed to handle high concurrency.
Each user:
- has their own private DM conversation
- runs their own AI generation
- does not block or affect other users
The system is designed to handle high concurrency.
Yes, Instagram (Meta) has platform-level limits for messages and automated
actions.
These limits depend on:
Important note:
Based on Apiway’s real usage history, we have not seen customers actually reach these limits during normal AI Virtual Try-On campaigns.
Try-ons are:
Because of this, Instagram treats them as natural interactions.
These limits depend on:
- account history
- trust score
- messaging behavior
Important note:
Based on Apiway’s real usage history, we have not seen customers actually reach these limits during normal AI Virtual Try-On campaigns.
Try-ons are:
- user-initiated
- 1:1 conversations
- not bulk or spam messaging
Because of this, Instagram treats them as natural interactions.
Yes.
When many users start a try-on simultaneously:
Queues help keep the system stable.
When many users start a try-on simultaneously:
- Requests are placed in a queue
- Users may wait slightly longer
- Results are still delivered
Queues help keep the system stable.
Yes. You can:
- pause individual flows
- disable automations
- turn try-ons off during nights, holidays, or off-season
If you cancel your subscription:
Your configurations are not immediately deleted, but the feature remains disabled until reactivation.
- Automations stop responding
- DM triggers stop working
- Try-on links become inactive
Your configurations are not immediately deleted, but the feature remains disabled until reactivation.
Marketing, Strategy & Analytics
You should actively promote it.
AI Virtual Try-On is still a new behaviour for most users.
If you don’t explain it clearly, many people won’t understand what to do.
Best results come when you:
AI Virtual Try-On is still a new behaviour for most users.
If you don’t explain it clearly, many people won’t understand what to do.
Best results come when you:
- Explain it visually (carousel, Reel, Story)
- Repeat the message several times
- Clearly say what action to take (“comment TRYON”, “send keyword”, etc.)
Keep it simple and visual.
Best formula:
Example:
“Comment TRYON and see this dress on yourself in DMs.”
Avoid long explanations — people don’t read.
Best formula:
- Show the garment
- Show a try-on result
- Explain in one sentence what to do
Example:
“Comment TRYON and see this dress on yourself in DMs.”
Avoid long explanations — people don’t read.
Yes — and you should use them.
Templates usually include:
Templates usually include:
- short caption with one CTA
- pinned comment explaining the keyword
- last slide in carousel with CTA
- Story with “Swipe / Comment to Try-On”
Both work, but for different goals.
Comment triggers
Ad (Click-to-DM) triggers
For most brands:
👉 start with comments, scale with ads.
Comment triggers
- Great for organic reach
- Strong engagement
- Good for the existing audience
Ad (Click-to-DM) triggers
- Better for predictable sales
- Works well with cold traffic
- Easier to scale
For most brands:
👉 start with comments, scale with ads.
Best-performing creatives:
Highly polished ads usually perform worse than simple, human-looking content.
- UGC-style videos
- Short screen recordings of try-on results
- “Before / After” style visuals
Highly polished ads usually perform worse than simple, human-looking content.
Use simple, obvious keywords.
Best examples:
Avoid long or complex keywords.
People should understand what to type instantly.
Best examples:
- “tryon”
- “fit”
- product name (e.g. “dress12”)
Avoid long or complex keywords.
People should understand what to type instantly.
Start with a curated selection.
Best candidates:
Best candidates:
- top-body clothing
- outerwear
- dresses that already sell well
Use links in the DM flow.
Track:
Track:
- clicks on “Buy Now” buttons
- purchases on your site
- conversion from DM → site
Yes.
You can add:
You can add:
- UTM parameters
- campaign tags
- custom links
- Google Analytics
- Shopify
- other analytics tools
Yes — indirectly.
Common strategies:
Common strategies:
- DM follow-ups
- retargeting ads to DM engagers
- email capture inside the flow
Both work.
For launches
For launches
- creates excitement
- drives engagement
- revives old posts
- increases conversion
Yes — this works very well.
You can:
You can:
- repost try-on results (with permission)
- show real user examples
- create “community try-on” Stories
Yes — and it boosts participation.
Examples:
Examples:
- “Best try-on of the week wins a discount”
- “Share your try-on and tag us”
- “Vote for the best look”
Indirectly — yes.
Why:
Why:
- more comments
- more DMs
- more saves and shares
Yes.
You can export or connect data such as:
You can export or connect data such as:
- try-on counts
- clicks
- engagement metrics
Yes. Common flow:
- user tries on
- user leaves email (optional)
- follow-up via email
AI Virtual Try-On is not just a gimmick.
It works best when you treat it as:
It works best when you treat it as:
- an engagement hook
- a DM conversion layer
- a bridge between content and checkout
Customer Reactions, Support & Psychology
From real usage, reactions usually split like this:
This distribution is normal and expected.
- ~70–80% WOW
- ~15–20% neutral
- ~5–10% negative
This distribution is normal and expected.
Acknowledge the feeling first.
Best response:
Best response:
- explain that AI Try-On is a visual simulation, not a perfect mirror
- suggest retrying with a better photo
- offer a second try
This is a psychological reaction, not just a technical one.
AI Try-On simulates the experience of:
“Looking at yourself in the mirror.”
Many people are uncomfortable with how they look — AI only exposes that.
Your role:
Never tell the customer they are “wrong”.
AI Try-On simulates the experience of:
“Looking at yourself in the mirror.”
Many people are uncomfortable with how they look — AI only exposes that.
Your role:
- stay calm and empathetic
- explain that AI doesn’t understand real body measurements
- encourage a retry or a different pose
Never tell the customer they are “wrong”.
Most customers blame the AI, not the brand — if you communicate
clearly.
Important:
When expectations are set correctly, backlash is rare.
Important:
- always position try-on as a preview / simulation
- avoid phrases like “perfect fit” or “100% accurate”
When expectations are set correctly, backlash is rare.
Agree with them — partially.
Best approach:
“You’re right — it’s not 100% accurate. It’s a visual preview to help you imagine the fit.”
People trust honesty more than hype.
Best approach:
“You’re right — it’s not 100% accurate. It’s a visual preview to help you imagine the fit.”
People trust honesty more than hype.
Photo rejection happens because of:
Best response:
Keep it short and friendly.
- AI safety systems
- poor lighting
- unusual poses
- suspected celebrity photos
Best response:
- explain briefly why it happens
- ask for a simple retry
- provide clear photo tips
Keep it short and friendly.
Tone matters more than technology.
Best practices:
Best practices:
- use casual language
- emojis (sparingly)
- playful CTAs (“Try it 👀”, “See yourself in it”)
- short messages
Don’t hide from them — respond publicly and calmly.
Good response structure:
This builds trust with future customers.
Good response structure:
- Thank them for the feedback
- Acknowledge the limitation
- Clarify that AI Try-On is a preview, not a guarantee
This builds trust with future customers.
Yes. Support teams should understand:
Train them to:
Empathy is more important than technical explanations.
- AI is not perfect
- retries are normal
- emotional reactions are common
Train them to:
- reassure, not debate
- guide, not defend
Empathy is more important than technical explanations.
Yes.Best practice:
Most users won’t need it, but it’s important for edge cases.
- add a “Talk to a human” or “Need help?” option
- allow manual DM takeover when needed
Most users won’t need it, but it’s important for edge cases.
Marketing, FOMO & Content Strategy
FOMO works extremely well with try-on.
Effective tactics:
Effective tactics:
- “Try today only”
- “First 100 people”
- “Exclusive for followers”
- “Only this week”
Best balance - 1 try-on post per 3–4 regular posts.
Too much - people get bored.
Too little - people forget it exists.
Too much - people get bored.
Too little - people forget it exists.
Yes, strongly.
It increases:
It increases:
- comments
- DM opens
- shares
- saves
Yes.
You do this by:
You do this by:
- creating separate flows per influencer
- using unique links or keywords
- tagging flows internally
- influencer ROI tracking
- performance-based payouts
- A/B testing creators
Yes — and it works extremely well.
Creators use try-on as:
Creators use try-on as:
- first-touch engagement
- “fun value” before selling
- trust builder
Not fully white-label yet.
Agencies can:
Agencies can:
- sell it as a service
- manage flows for clients
- charge retainers
- Apiway branding may still appear
- full white-label requires custom setup
Each Instagram account must be connected separately (Meta rule).
In practice:
In practice:
- one Apiway workspace
- multiple connected accounts
- clear naming + structure
Funnel, Drop-Offs & Retargeting
Typically:
- 20–30% drop-off
- they don’t have a photo ready
- privacy hesitation
- distraction
Best methods:
- DM reminder (“Want to try it now?”)
- Story reminders
- Ads to engagers
Yes.
Examples:
Examples:
- “Still curious how it looks on you?”
- “Want to try with another photo?”
Best windows:
- 30–60 minutes
- or on the next day
Feedback, Optimization & Brand Control
Yes.
You can ask:
You can ask:
- “How does it look?”
- “Would you wear this?”
- emoji reactions
- conversion insight
- product feedback
- content ideas
Yes.
You can control:
You can control:
- lighting
- realism level
- background mood
- fashion style
Yes.
Test:
Test:
- short vs long
- playful vs serious
- emoji vs no emoji
GA4 can’t see DMs directly.
But:
But:
- UTMs work
- link tracking works
- purchase attribution works
Use:
- UTMs
- discount codes
- post-try-on reminders
Website & Embedding
Yes, very common.
Use:
Use:
- Linktree-style pages
- Landing pages
- FAQs
Currently:
- DM-first experience
- website redirects possible
- full widget = future roadmap
Product Categories & AI Limits
Works with limitations.
Problems:
Problems:
- transparency
- safety filters
- pose sensitivity
Technically possible, but:
- Stricter safety rules
- A higher rejection rate
Partially.
Nano Banana 2.0:
Nano Banana 2.0:
- much better at layering
- still not perfect
Language, Devices & Platforms
Yes.
You can fully localize:
You can fully localize:
- DM messages
- Photo instructions
- CTAs
Yes.
Creating separate flows per country or market works best.
This allows you to customize:
Creating separate flows per country or market works best.
This allows you to customize:
- language
- tone
- offers
- links
Mobile = best
Desktop = limited
AI Virtual Try-On works best inside the Instagram mobile app.
On desktop Instagram, DM buttons are not clickable.
This is a Meta / Instagram limitation, not an Apiway issue.
Because of this:
Desktop = limited
AI Virtual Try-On works best inside the Instagram mobile app.
On desktop Instagram, DM buttons are not clickable.
This is a Meta / Instagram limitation, not an Apiway issue.
Because of this:
- Users cannot properly interact with try-on buttons on desktop
- The full try-on experience is not available
Yes.
On desktop Instagram:
On desktop Instagram:
- DM buttons do not work
- Some interactions are limited or broken
- The experience is incomplete
Trusted by 5,000+ Brands & Ads Experts
Worldwide
Official Meta Business Partner
and Tech Provider
We Use the Official Meta Marketing API for
Security
