Photo to Video is a browser-based AI image to video generator from Joseph Works Limited that animates a still photo with third-party models such as Seedance, Kling, Wan, Hailuo and Vidu, charging credits per generation instead of a subscription.
Used this tool? Rate it
Used this tool? Rate it
Photo to Video is a browser-based AI image to video generator at photo2video.ai: upload a photo, add an optional prompt, choose a model and generate a short clip. The product's own pitch is to turn a photo into video with AI and add a prompt for smooth, cinematic motion.
Instead of a single model, the site's stated approach is to put leading video models in one generator, so you can switch models per shot, compare results, and pay per video with no subscription. Models such as Seedance, Kling, Wan, Veo and Sora are all one click apart in the same generator and on the same credits.
The service is run by Joseph Works Limited, whose Terms of Service cover the Photo to Video website at photo2video.ai and its browser-based application. The company's own site lists Joseph Works Limited as registered in Hong Kong SAR under company number 80304232, incorporated on 29 April 2026. The App Store listing "Photo to Video: AI Video Maker" is sold by Joseph Works Limited and names photo2video.ai as its seller website, tying the web product and the iOS app to the same operator.
"Photo to Video" is also a generic phrase used by several unrelated sites and apps; this page covers only the product at photo2video.ai.
The model pages each answer whether they are the official site for that model, and the answer is no. The Wan page says Photo to Video is an independent product that runs Wan, Alibaba's video model, through official API partners and is not affiliated with or endorsed by Alibaba.
The homepage model strip lists Seedance, Wan, Seedance 2, MiniMax H3, Kling, Vidu, Hailuo, Happy Horse, Agnes, LTX 2.5, Imagine, Veo and Sora 2. Each model has its own tiers, lengths and resolutions, so the model choice decides most of what a clip can do and what it costs. Seedance 2.5 renders up to 30 seconds as one continuous take, where Seedance 2.0 stops at 15.
Generation speed depends on the model: faster options can finish in seconds, while most generations take about one to two minutes.
Reference-image limits are model-specific, and the pages give different figures. The homepage says up to eight reference images can be uploaded so the model holds shape, color and composition from frame to frame. The LTX 2.5 page says Seedance 2.5 takes up to 30 reference images, Kling and Vidu take seven, and LTX 2.5 takes none.
A built-in AI photo editor offers Seedream 4.5, Seedream 5 Pro, WAN 2.7, WAN 2.7 Pro and Hunyuan Image 3.0 Instruct, picked from a menu like the video models. One tap sends the edited image to the video generator, so a retouched photo does not need a second upload.
The App Store description lists a Transition Mode that animates between uploaded first and last frames, Motion Control that copies poses and actions from a reference video onto a character image, and a Video Extend mode that adds 4 to 12 seconds to an existing video. The web pages checked this round do not describe these modes.
The Seedance 2.5 page describes a draft-then-rerun pattern: draft the shot at 480p or on Turbo first, then re-run the version you like at 4K, since the prompt and image carry over.
A pattern runs through these examples: they start from a single existing image. Clips of a recognisable person also bring in the consent rules under Limitations.
It fits:
It is a weaker fit for:
Photo to Video sells credits, not plans. The terms describe one-time, non-renewing consumable credit packages, with the credits charged depending on the model, duration, resolution, audio setting and other options shown before generation.
New accounts start with free credits: the Seedance 2.5 page cites 50 credits for a new signed-in account. The terms add that free credits are granted once per account rather than once per sign-in, and that free or promotional credits may be changed or withdrawn where permitted by law. The model pages repeat that there is no card up front and that the exact price appears on the Make Video button before anything is spent.
Credit-pack prices are not published on the public pages checked this round. According to the terms, prices, currency, taxes and the number of credits included are shown before you confirm a purchase. A directory listing submitted for the product shows pricing as freemium from $4.99; that figure could not be matched to a price on photo2video.ai this round.
Figures are the examples each model page publishes, not a full rate card.
| Model | Cheapest example on the page | Credits | Top example on the page | Credits |
|---|---|---|---|---|
| Hailuo | 6 s at 512P | 13 | 6 s on 2.3 Pro at 1080P | 59 |
| MiniMax H3 | 5 s on Fast at 480P | 15 | 5 s on Pro at 2K | 126 |
| Vidu | 5 s on Fast at 540P | 30 | 5 s on Pro at 4K | 113 |
| Kling | 5 s on 2.6 Standard | 32 | 5 s on 3.0 4K | 378 |
| Wan | 5 s at 720p, sound off | 17 | 5 s at 720p with sound | 33 |
| Seedance 2.0 Mini | 5 s at 480p | 36 | 5 s at 1080p | 180 |
| Grok Imagine | 5 s, any resolution | 38 | 15 s, any resolution | 113 |
| LTX 2.5 | 5 s at 720p | 90 | 5 s at native 4K | 297 |
| Seedance 2.5 | 5 s at 480p | 162 | 5 s at native 4K | 1,620 |
Two consequences are easy to miss. Free credits do not reach every model: the cheapest LTX clip is 90 credits because LTX 2.5 starts at 720p and has no 480p tier. And the site states plainly that unlimited free Seedance 2.0 does not exist, because every clip costs real GPU time to render.
The AI image editor is priced per image: 4 credits on WAN 2.7, 5 on Seedream 4.5, 13 on Seedream 5 Pro at HD, 14 on WAN 2.7 Pro and 17 on Hunyuan 3.0 Instruct. An edit the model declines is refunded, according to the editor page.
Two independent discovery routes point in different directions. One launch directory lists Pollo AI, Higgsfield and Runway as alternatives. An AI tool directory instead groups it with photo-animation tools such as Kadapt, Ebsynth, Deep Nostalgia and AnimateMyPic.
The other route is the model makers themselves, since Photo to Video reaches models such as LTX 2.5 from Lightricks through official API partners rather than owning them. The trade-off is account overhead: the Grok Imagine page notes that using the model here needs no xAI account, no X Premium and no Grok subscription.
An anonymous account is fragile. The terms say that losing access to your sign-in provider, or clearing browser data or changing devices on an anonymous account, may prevent recovery of its history or credits.
The privacy policy says content submitted for generation is screened by automated content-safety systems as part of processing. The AI image editor page says the browser version runs no safety checker of its own, so nothing sits between your prompt and the model. That page reports that over the past month Seedream declined about 4% of edits and WAN 2.7 about 5%, while Hunyuan 3.0 Instruct declined none of the 3,500 edits it ran. The pages checked this round do not reconcile the two statements.
The prohibited-use list bans non-consensual intimate content, sexual content involving minors and exploitative content, as well as misleading media for fraud, election interference, identity abuse or unlawful surveillance.
As of September 23, 2026, no independent media review of photo2video.ai was found in the news and technology channels searched, and the consumer review platform checked had no profile for it.
Only with the rights and permissions needed. The terms make the uploader responsible for permission from people depicted when required, and they ban non-consensual intimate content and media made for identity abuse.
The privacy policy and terms do not say. They do state that uploads go to the selected AI provider and are not used for advertising.
Generally no. A result that misses a subjective expectation does not qualify on its own, though credits for generations that failed through a technical error may be restored.
