Prompt & Done · SECRETSITES
3 free AI sites, no account: what we typed and what we got
We ran all three ourselves on a phone-sized browser, logged out, no card. One sentence became a photo, the photo became a moving clip, and a second sentence became a video with no photo at all.
Tested 7 Nov 2026. Free tiers and limits change; check the site before you rely on it. No affiliate links.
1 duck.ai: OpenAI's image model, no account
DuckDuckGo's private AI chat. On the free plan, without logging in, it offers GPT-6 Luna, GPT-5.4 mini, Claude Haiku 5.5, Mistral Small 4, gpt-oss 120B and Gemma 4 31B, plus Create & Edit Images. Our image came back labelled GPT Image 2.5 in 19 seconds.
Our prompt: A tiny orange robot running a ramen stall in a neon night market at night, rain, cinematic photo, vertical 9:16
Catch: the free plan has usage limits; the bigger models (Claude Sonnet, Opus, GPT-5.6) need a subscription. Chats are anonymized; DuckDuckGo says they aren't used for AI training.
2 vheer.com: make a photo move
Open Image to Video, drop in a picture, describe the motion, press Generate. We used the duck.ai photo and got a 4.7-second clip (push-in, rising steam, a little robot movement) in about 2.5 minutes. No signup, no watermark on our export.
Our motion prompt: The little robot lifts the noodles with chopsticks and smiles, steam rising, rain falling, neon flickers
Catch: motion is subtle and generation took 2.5 to 4.5 minutes; accept the privacy banner first or Generate may not start.
3 raphael.app: a video from one sentence
In AI Video, pick Text to Video and 9:16, type a sentence. The free model (MiniMax H3 Turbo) made our 4-second 480p clip with no sign-up.
Our prompt: A tiny orange robot chef flips a glowing bowl of ramen in a rainy neon night market, steam rising, cinematic, slow push-in
Catch: the free queue took about 13 minutes, the page warns results are temporary (keep the tab open and download), and higher-quality models are paid.