How it works

Four steps

1. Sign in with an email code (your account is created on first sign-in, no password). You can also use Google.
2. On the Credits page, enter the code your reseller gave you; the credits arrive immediately and work for all five tools.
3. Create a job and pick a tool. Xiaohongshu posts: a topic or a pasted reference post, number of posts (1–30), images per post (1–4), a style, and optional brand and banned words. Explainer videos: a one-line topic or attached documents (either is enough), a length (30 seconds, 1 minute, 5 minutes or 10 minutes), a language, portrait or landscape, a style, a voice, and whether to burn in subtitles. One-click images: a one-line idea and 0–3 reference images, an optional free AI plan you can edit, the number of images (1–12) and a shape (3:4, 1:1, 4:3). Product promo videos: the product name and 1–5 selling points, 3–12 product photos (straight from a finished image set, or uploaded), a length (15 or 30 seconds), portrait or landscape, a template and music. Software promo videos: your website or a few lines about the product (at least one), 0–8 product screenshots, a length (15, 30 or 60 seconds), landscape or portrait, a voice-over and music; AI first proposes three directions for free, and you pick one and edit the on-screen words before generating. You see the estimated credits before submitting.
4. The job page shows progress. For posts, preview each one and redo any image you don't like; for a video, play it right on the job page and copy the storyboard and narration; for images, see the set laid out by shot and redo any single one; for product and software promo videos, play the film right on the job page and edit its text for free, and on a software promo you can also point at a shot to have it redrawn. Then download the ZIP.

How posts are made

• First, as many distinct angles as there are posts, so the posts don't all read the same.
• Each post gets a title (up to 20 characters), copy (300–600 characters), 5–8 tags, and a scene description for each image.
• Images are generated from those descriptions, without text in the picture: AI-rendered Chinese text is often wrong, so add text when you lay out the post.

How videos are made

An AI director splits the topic or documents into pages by length and writes each page's visuals and narration; each page gets one image and, if you picked a voice, a voice-over; an AI artist animates each page, and the result is rendered to mp4. On-screen text is typeset by a program, so there are no garbled characters. The Explainer videos page covers each step, the styles and the files you get.

How images are made

AI first looks at your references and reads the brief to draft an editable plan (free, up to 30 times per person per 24 hours), or you skip it and generate directly. Once you confirm, the scene sets out a list of shots, for a product set: white-background main, selling-point hero, selling point, in use… Images are generated one by one, and with references the subject stays true to the original. The One-click images page lists every scene's shots and when text goes on an image.

How product promo videos are made

AI looks at the photos and writes the copy: an opening line, the photo order and captions, the wording of each selling point, a closing line and the post copy; a number you didn't write sends it back for a rewrite, and if two rounds both fail your own words are used as is. The program lays out the picture by template (photo wall, quick cuts, spotlight): it fixes the music's tempo first, every page is a whole number of beats and every cut lands on a beat, with colours from your brand colour or the main colour of the photos. Music and sound effects at the same tempo are added and it's rendered to mp4. The Product promo videos page shows the three templates and the files you get.

How software promo videos are made

With a website, it first opens the home page and up to two more pages on the same site, takes screenshots, reads the text and picks up the brand colours and logo; if the site won't open, it goes by your description and screenshots alone. Once AI has read the product, it proposes three directions with shot lists for free (up to 10 times per person per 24 hours); you pick one, edit the words and delete shots. When generating, the artist draws shot by shot, rebuilding interface pieces from your screenshots, and can only use the words in the plan; the program sends failing drafts back, up to 3 per shot, and if that still fails a program-made fallback shot takes its place. Music follows the direction's tempo, cuts land on the beat, and it's rendered to mp4. The Software promo videos page covers each step.

When something fails

If the generation service is busy, every kind of job waits and continues automatically; no need to resubmit.

Xiaohongshu posts are charged step by step: a step rejected by content review, or still failing after retries, is not charged and the rest is delivered; rejected images can be redone one at a time.

Explainer videos are charged as a whole: only once the video is made. If any step ultimately fails (rejected by content review, or still failing after retries), no video is produced and the whole hold is released.

One-click images are charged per image: each one is charged when it succeeds, one that fails or is rejected is refunded and can be redone. Planning is free; if it doesn't produce a plan, try again or fill one in yourself.

Product promo videos are charged as a whole: only once the video is made; if it isn't, every held credit is returned. Editing text afterwards is free.

Software promo videos are charged as a whole: planning is free; you're only charged once the video is made, and if it isn't, every held credit is returned. If a shot fails all 3 drafts, a program-made fallback shot takes its place and the film still comes out. Point to fix redraws one shot in up to 2 drafts and is charged only once it's in the film; if both drafts fail, nothing is charged and the film stays as it was.