Photo AI
Photo AI trains a personal AI model from 5 to 20 selfies, then generates photorealistic photos and videos of you or of a fully synthetic influencer. For creators, marketers and fashion brands, on a paid subscription from $19 monthly.
What is Photo AI?
Photo AI is a web product that generates photorealistic photos and videos of people. It presents itself as the first AI Photographer in the world and is the declared successor of Avatar AI, the original viral AI avatar generator that launched the global trend back in 2022; the footer still reads "Formerly known as Avatar AI". The service started on photoai.io, first captured by the Wayback Machine on 13 December 2022, and is served today from photoai.com.
The principle is fine-tuning. You upload 5 to 20 varied selfies, the service trains a personal AI model in roughly one minute, then generates photos and videos of that subject on demand. If you prefer not to appear at all, you can instead build a fully synthetic influencer by choosing origin, body type, hairstyle and eye colour.
Generation runs on a proprietary pipeline called Hyper Realism. Google's Nano Banana Pro can be selected as an alternative model and ByteDance's Seedance 2.5 handles video; the publisher describes a technical history running from Stable Diffusion to Flux. More than 100 themed photo packs cover dating, professional portraits, fashion, travel, social media, retro, fitness, events and parties.
Around generation sits an editing suite: Magic Edit in natural language, cropping, zoom out and uncrop, relighting, photo combination and an upscaler. Video goes from image to clip, including talking videos with a voice, and a multi-track editor with export runs in the browser. Photo AI also produces T-pose 3D models intended for Mixamo rigging, and clothing try-on, single or batch, aimed at fashion and e-commerce. A separate section offers 18 free side tools with no account: background remover, upscaler, watermark remover, face swap, colouriser, restoration, HEIC converters, compressor, collage maker and a photo location anonymiser. The home page counter shows 30,528,382 photos generated.
Behind the product is one person. Pieter Levels runs it as a 100%-owned independent business without investors, with no employees and claimed margins of 75 to 90%. Two scope limits matter before signing up: access is web only, the iOS app having existed and been withdrawn, and there is no public API.
What it does
- Train a personal AI model on 5 to 20 varied selfies in about one minute.
- Generate photos from a free-text prompt or by launching one of more than 100 themed photo packs.
- Turn any photo into a video with a single Make video action.
- Edit images in plain English with Magic Edit: add, remove or change the setting or the style.
- Try clothes on your model, one garment at a time or in batch.
- Generate hundreds of photos at once from a list of prompts separated by |||.
- Assemble a complete film in the multi-track video editor that runs in the browser.
When to use Photo AI / When not to
A quick filter to help you decide if Photo AI is the right fit.
When to use Photo AI
- Content creators and influencers who need a steady stream of social media photos and videos without booking a shoot, with dedicated Instagram, Instagram Carousel and TikTok Slideshow packs.
- Professionals who need current portraits: founders and CEOs, lawyers, doctors, teachers, therapists, real estate agents and actors, with a LinkedIn headshot pack among the options.
- Singles building dating profiles, served by dedicated Tinder, Hinge, Bumble and Badoo photo packs.
- Fashion and e-commerce teams, explicitly Shopify sellers, fashion brands and indie designers, running batch clothing try-ons and product videos presented by an AI model.
- Marketers building a fully synthetic AI influencer to monetise, with origin, body type, hairstyle and eye colour configured from scratch, and anyone replacing a photographer billed in the hundreds or thousands.
When not to use Photo AI
- Anyone under 18: the terms require users to be 18, or the local age of majority if it is higher.
- Anyone looking for nudity or sexually explicit imagery: the publisher states that Photo AI is not a nudifier or undressing application, and uploads or generations involving minors are monitored and reported to the authorities.
- Anyone wanting images of other people without their explicit written consent, images of politically exposed persons, or deceptive deepfakes for political campaigning, advocacy, misinformation or scams, all of which the terms prohibit.
- Developers and mobile-first users: there is no public API documentation, no programmatic access, and no mobile app since the iOS app was withdrawn.
- Procurement-driven teams that need a data processing agreement, a subprocessor list or a documented hosting region: none of the three is published.
How to use Photo AI
A typical end-to-end flow, from setup to results.
- Sign up with an email address or a Google account; accepting the terms of service is required at registration.
- Click Create new AI model and fill in the model name, gender, age, ethnicity and eye colour.
- Upload around 20 varied photos, mixing angles, lighting, backgrounds and expressions, and avoid filters, sunglasses, Instagram screenshots, group shots and cropped faces.
- Let training run automatically for about one minute; a notification email confirms when the model is ready.
- Collect the free auto-generated photos that come with each model created, 48 per model on Pro, Max and Ultra.
- Generate images by writing a prompt, launching a photo pack, or copying an existing photo with copycat; the home page announces about 20 seconds per photo.
- Hover a photo and click Make video to turn it into a clip.
- Hover a photo, click Magic Edit and describe the change in one sentence; Magic Edit does not use your personal model, so run Remix to restore the likeness.
- Open the Try on clothes panel and upload several photos of garments, flat or worn; for volume, paste a list of prompts separated by ||| and choose 1, 2, 4, 8, 16 or 32 photos per prompt on the Premium and Ultra plans.
- Manage billing and cancellation at photoai.com/billing through the Stripe customer portal using your purchase email, and report bugs or request features on ideasandbugs.com with the same email as your account.
Pros & Cons
Pros
- Likeness is the point: the model is trained on your real photos, the argument the publisher makes against general-purpose generators that cannot train on them.
- Fast throughout: about one minute to train a model and roughly 20 seconds per generated photo.
- Unusually broad coverage for a single subscription, spanning photo, video, editing, 3D and clothing try-on.
- More than 100 ready-made packs remove the need to write prompts, and each model created on Pro, Max or Ultra comes with 48 free auto-generated photos.
- 18 side tools are genuinely free, with no account, no email and no watermark.
- Annual billing cuts the price roughly in half, with 6+ months free and an entry point at $9 per month.
- An independent publisher with no investors, who publishes his business model, states he is profitable, announces bug fixes within hours and displays testimonials marked Verified purchase.
Cons
- No free plan for the generator: every use of it goes through a paid subscription.
- Refunds are highly restricted, available only if no model has been created, fewer than 20 photos have been generated and the account was not opened through a referral link, and a waiver of the 14-day withdrawal right is imposed at purchase.
- No postal address, no incorporated legal entity and no human contact email are published; support runs through a third-party bug board, ideasandbugs.com, and the community, in a one-person business with no employees.
- No data processing agreement, no subprocessor list, no documented hosting country, and no documented way to exclude your data from model training.
- A perpetual, irrevocable and sublicensable licence is granted to the publisher over prompts, input images and Assets, and it survives termination.
- Singapore governing law and jurisdiction, mandatory individual arbitration, and waivers of class actions and jury trial.
- No mobile app since the iOS app was withdrawn and no API; the commercial use licence is absent from Starter, and batch generation, batch try-on, video and the upscaler are reserved for higher plans.
Pricing & Plans
There is no free plan for the generator: every use of it requires a paid subscription, and the publisher explicitly justifies the absence of a free tier. The lowest price point is USD 19 per month with the Starter plan, or USD 99 per year, equivalent to USD 9 per month, on annual billing. Four plans are offered, Starter, Pro at USD 49, Max at USD 99 and Ultra at USD 199 per month, with monthly credit allowances of 50, 1,000, 3,000 and 10,000 respectively, renewed every month including on annual billing. Generation is metered in credits: 2 for a Hyper Realism photo, 5 for a Nano Banana Pro photo, 1 for an import, 100 for a video with sound and 25 for a 3D model. Payment is processed by Stripe and taxes are not included. Cancellation is possible at any time through the Stripe portal at photoai.com/billing, without refund of amounts already billed, and refunds themselves are limited to accounts with no model created, fewer than 20 photos generated and no referral sign-up. There are no coupon codes: annual billing is the only discount. Separately, 18 side tools are entirely free, with no account and no email required.
- 50 credits per month
- Nano Banana Pro and Hyper Realism
- low photo quality
- low likeness
- 1 photo at a time
- slow processing
- personal use only
- no free auto-generated photos.
- 1
- 000 credits per month
- medium quality and likeness
- 4 photos in parallel
- photo imports
- custom prompts
- remix
- commercial use licence
- 3
- 000 credits per month
- high quality and high likeness
- 8 photos in parallel
- editing
- cropping
- zoom out
- AI videos
- 10
- 000 credits per month
- ultra quality and likeness
- 16 photos in parallel
- unlimited photo storage
- priority response times
- 2
- 400 free photos in total.
- Annual commitment was advertised at the time of collection with a 6 MONTHS FREE countdown offer.
Data, GDPR & hosting
A consolidated view of how Photo AI handles your data.
GDPR overview
Photo AI publishes a single document combining terms of service and privacy policy, effective and last updated 10 October 2024. It grants GDPR and CCPA rights to all users regardless of location, and cites contract performance, legitimate interest in security, reliability and attribution, and withdrawable consent as legal bases. EEA, UK and Swiss users are told they may rectify, erase, object, port their data, withdraw consent and lodge a complaint with a supervisory authority; Californian users get deletion, an opt-out of sale, authorised agents and identity verification. The EU AI Act is cited regarding non-consensual intimate imagery. The gaps are equally concrete: no Article 27 EU representative, no data protection officer, no postal address, no privacy contact email (the only address on the site is no-reply@photoai.com), no data processing agreement and no subprocessor list. Governing law and jurisdiction are Singapore.
Who owns the data?
Under the combined terms and privacy policy in force since 10 October 2024, you own the Assets you create: Photo AI assigns you all rights, title and interest in them for personal or commercial use, although the commercial licence starts at the Pro plan, Starter output being for personal use only. In exchange you grant Photo AI and Interior AI, the sibling product that shares this legal document, together with their successors and assigns, a perpetual, worldwide, non-exclusive, sublicensable, no-charge, royalty-free and irrevocable copyright licence over your prompts, the images you upload and the Assets produced, covering reproduction, adaptation, publication, display and distribution. That licence survives termination. Assets you make public inside the service are additionally cross-licensed to other users.
Reuse rights
The same document, effective 10 October 2024, lists the data collected: name, email, payment information including card numbers and bank details, the photos and files you upload, IP address and device identifiers, geolocation depending on your settings, usage data and inferences, plus information received from partners, connected social accounts and service providers. Stated purposes are to provide the service, improve and develop it, communicate and deliver support. Data is shared with your consent, with third-party providers held to the same level of protection, to meet legal obligations, to prevent harm and in a corporate restructuring; the publisher states it does not sell personal information and has not done so in the past twelve months. Uploaded and generated content is monitored, and content involving minors is reported. On the output side, users can reuse their Assets without asking permission, commercially from the Pro plan upwards and for personal use only on Starter. The training link is only implicit: the licence taken on prompts, input images and Assets expressly covers improving the services, and the publisher says it keeps improving its pipeline through reinforcement learning from human feedback by users. No opt-out from that use is documented.
Data retention & training
Hosting summary
Photo AI names no hosting country and no hosting region. The legal document, effective 10 October 2024, states only that as part of normal operations information may be stored on computers in countries outside the user's home country, and that by providing information the user consents to that transfer. The publisher undertakes to comply with applicable law wherever data is stored, and describes physical, organisational and technical safeguards along with breach notification. No data processing agreement and no subprocessor list are published, and no Article 27 EU representative is designated. Technical signals are indicative only: the domain sits behind Cloudflare, autonomous system AS13335, and media assets are served from a bucket named r2-us-west.photoai.com. Jurisdiction is explicit where hosting is not, the terms placing governing law, jurisdiction and arbitration in Singapore. Organisations with data residency requirements should treat the hosting location as undocumented.
Things to keep in mind
Risks and trade-offs to weigh before adopting Photo AI.
- Buying waives the 14-day withdrawal right, and refunds become effectively unreachable as soon as a model is created or 20 photos are generated.
- Subscriptions renew automatically and must be cancelled before the renewal date; the publisher reports that double subscriptions opened on two different email addresses are a frequent case.
- The perpetual, irrevocable and sublicensable licence granted to the publisher over prompts, input images and Assets survives termination, and no training opt-out is documented while reinforcement learning from user feedback is claimed.
- Disputes fall under Singapore law with AAA arbitration in Singapore and waivers of class actions and jury trial; the 30-day arbitration opt-out must be sent to a Notice Address that is published nowhere, and any claim must be brought within one year.
- The same legal document states two different liability caps, USD 500 in one clause and USD 100 in another.
- Photorealistic portraits that were never taken can distort self-image and mislead audiences on dating apps, job profiles and social media; the more convincing the likeness, the easier it becomes to rely on a face that does not exist.
- Generating images of other people requires explicit written consent and the legal risk sits with the user; third-party applications trading on the Photo AI name are flagged by the publisher as counterfeits.
Setup & Integrations
Technical difficulty
Very low. Nothing to install, no hardware, no code: everything happens in the browser. Sign up with an email or Google account, upload selfies, and training runs automatically in about a minute. Pre-written packs remove the need to write prompts. The only real learning curve is selfie quality, plus prompt writing for advanced work; batch generation with the ||| separator and JSON prompts with Nano Banana Pro are more technical. The publisher contrasts this with the local alternative, an RTX 3090-class GPU with Python, PyTorch CUDA, diffusers and DreamBooth, which he calls complex and very slow.
Deployment
Integrations
Supported languages
Behind Photo AI
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Photo AI.
Frequently asked questions
How many photos do I need to upload to train a model?
How long does training take?
Is there a free plan?
Which AI models does Photo AI use?
Can I use the generated images commercially?
Is there a mobile app or a public API?
Can I get a refund, and how do I cancel?
Can I delete my data?
Who can appear in the photos?
Who runs Photo AI and how do I get support?
Should you pick Photo AI?
Photo AI does the one thing general-purpose generators still fail at: it trains on your actual photos, so the person in the output looks like the person who signed up. That likeness, obtained after about a minute of training, is the reason to consider it. Around it sits a functional range that is unusual for a product built and run by a single person, from photo packs and natural language editing to image-to-video, a multi-track video editor, 3D models for rigging and clothing try-on, all inside one subscription.
The reservations concern governance more than output. Legal transparency is minimal: no postal address, no incorporated company name, no human contact email, no data processing agreement, no subprocessor list, no named hosting country, and Singapore law with mandatory individual arbitration. The licence granted in return is broad, being perpetual, irrevocable and sublicensable over prompts, uploaded images and generated Assets, and it survives termination, with no documented training opt-out while the publisher openly describes improving its pipeline with reinforcement learning from user feedback.
Commercially, expect no free tier for the generator and refunds that stop being available as soon as a model is created or 20 photos are generated. The 18 free side tools are real, but they are not the product.
Photo AI suits creators, marketers, fashion and e-commerce brands, and professionals who need portraits on demand and are comfortable with a fast-moving, one-person publisher. Teams with procurement, DPA or data residency requirements should look elsewhere, and anyone planning to generate images of people other than themselves should read the consent rules first. Judge it on likeness and breadth, and go in with open eyes on the legal terms.
- Choosing a selection results in a full page refresh.
- Opens in a new window.