Video AI
Describe your product in one sentence. The agent writes a shot-by-shot storyboard, renders every shot from your product photo and stitches a finished video with voiceover and subtitles.
A short chat replaces the brief, the storyboard meeting and the edit.
Share the product, its selling points and who it's for in one or two sentences. The agent fills in the rest and asks only what it needs.
Upload one clean photo. Every shot's first frame is generated from it, so the packaging, logo and colors stay the same across the video.
The agent writes a shot-by-shot storyboard with visuals, voiceover lines and timing. Edit any shot directly or ask for changes in the chat.
Confirm the first-frame price and see every shot as a still before any video is made. Redo the ones you don't like, then confirm the video price to render each shot with its voiceover.
Shots are stitched in order, the voiceover lands on each shot and subtitles are added from the exact script. Download it or find it in My Creations.
Most AI storyboard generators stop at sketches. Here every shot of the storyboard is a working row: the visual, the voiceover line, the camera move and the length, plus its own first frame and video. Change one shot without touching the others, reorder shots, or split a long line into two. Each shot is drawn as a still first frame before any video is made, so you only spend credits on video once the shots look right.
Each shot starts from a first frame built on your product photo, then becomes its own clip. You confirm the price twice: once for the first frames, and once for the videos after you've checked them. When every clip is ready, the agent stitches them in order at one resolution and one aspect ratio, so nothing gets stretched or squashed between shots.
When one action is split across two shots, the agent marks the second shot to continue from the first: the first shot's video ends on the second shot's first frame, so the cut doesn't jump. That first frame is drawn from the person in the shot before to keep the face and hair, and the outfit is written the same way in both shots, while your product photo keeps the packaging and colors the same in every shot. Add a photo of a person to keep the same face throughout the video. If your product has a brand name or other text printed on it, each first frame gets an automatic fix that rewrites the lettering to your exact spelling, so garbled text is far less likely. Need one clip that follows a reference image? Try Reference to Video.
Write the script yourself or let the agent draft it. Every line is voiced first and timed to the shot, so the clip is as long as the line needs. Play a sample of each voice before you pick one. Subtitles come from the same script — brand names and numbers appear exactly as written, not as a speech-to-text guess.
Single generations top out at a few seconds. The agent plans a sequence of shots and joins them into one continuous video, so a 20–30 second ad with a hook, the key benefits and a call to action comes out of one conversation. When one action runs across two shots, the second shot picks up where the first ends, so the video plays as one piece instead of separate clips.
Turn one product photo and a few selling points into a finished multi-shot product video. The agent works from your selling points and audience, plans the shots around them, draws every first frame from your photo so the packaging and colors stay the same, fixes the brand name and other printed text to match your spelling, and ends on your call to action. For a single-shot ad with a spoken line, or a UGC-style ad, use the AI Ad Generator.
The same storyboard-first workflow suits short story videos: plan the beats in chat, keep each scene editable, and stitch the result. Product ads come first; more story formats are on the way.
Answers about storyboards, multi-shot consistency, editing, voiceover and pricing.
It's a chat assistant that plans and produces a video for you. Instead of writing one prompt per clip, you describe the goal; the agent writes the storyboard, generates each shot and assembles the final video, asking you to confirm the price before it spends any credits.
A storyboard can have up to 10 shots, and each shot is usually 4 to 8 seconds, so most videos land between 15 and 60 seconds. Product ads work best around 20 to 30 seconds.
Yes. A storyboard can have up to 10 shots, and every shot's first frame is generated from your product photo, so the product looks the same throughout. Add a photo of a person to keep the same face in every shot. When one action is split across two shots, the second shot continues from the first with the same person and outfit, and the first shot's video ends on the second shot's first frame, so the cut doesn't jump.
When two shots show one continuous action, the agent describes the person identically in both, and the second shot's first frame is drawn using the person from the first shot's frame (face, hair and what they wear). If the second shot doesn't describe the clothes, the outfit from the first shot is carried over. If you upload a photo of the person, every first frame uses it. You see all first frames before any video is made, so you can redo a shot that drifted.
Yes. Every shot has its own card: change the visual, the voiceover line or the prompts, then regenerate only that shot. You confirm the price for that shot before it runs.
Yes. Generation happens in two steps. First you confirm a price for the first frames only, one still image per shot. Redo any frame you don't like or change the shot's description. When they look right, the agent prices the videos and voiceovers, and nothing is spent on video until you confirm that second price.
Your product photo is the reference for every first frame, so the logo, packaging and colors follow it. If the product has a brand name or other text printed on it, the agent notes the exact spelling (you can also tell it), and every first frame gets an automatic pass that rewrites the lettering to match. That costs 2 credits per shot and is included in the first-frame price. Very small text or text partly hidden by a hand can still come out wrong; use Fix product text on that shot to try again for 2 credits.
Yes. Each shot's line is voiced, the shot length follows the voiceover, and subtitles are burned in from the exact script. You can switch subtitles off and choose from six voices, playing a sample of each first. Voiceover and on-screen text are English for now.
Yes. There's a play button next to each voice in the list and next to the voice picker. It plays a short sample for free, so you can compare the six voices before the agent records your lines.
Yes, and you should: every shot's first frame is generated from your product photo so the product looks the same throughout. You can also add a photo of a person to feature.
Chatting and writing the storyboard are free within a daily message limit. Credits are spent in two steps: first the first frames (a few credits per shot), then, once you're happy with them, the videos and voiceovers, priced by the video model and each shot's length. You see and confirm each total before anything runs. Stitching and subtitles are free, and failed steps are refunded automatically.
The AI Ad Generator makes one clip in a single pass, 1 to 20 seconds, with the spoken line, sound and motion generated together, which suits a quick single-shot or UGC-style ad. The AI Video Agent plans a multi-shot video: it writes a storyboard, lets you check every first frame, renders each shot, then stitches them with a timed voiceover and subtitles from your script.
Click Projects at the top of the chat to reopen a recent project with its storyboard, first frames and videos. Finished videos also appear in My Creations with a Video Agent label and an Open project button that takes you back to the project. First frames, single-shot clips and voiceovers stay inside the project and don't fill up My Creations.
Start with one sentence. The agent handles the storyboard, the shots and the edit.
Start with the agent