Generate coherent AI videos up to 30 seconds from text, images, or multimodal references with stable subjects, scenes, and style.
Seedance Mode
For dialogue, put the spoken words in double quotes. In reference mode, address files as @Image1, @Video1, and @Audio1.
Choose any whole-second duration from 4 to 30 seconds.
Auto lets the model choose the aspect ratio.
Generate synchronized ambience, effects, and dialogue audio at no additional token cost.
Sign in to spend tokens
Prompt is required
Your generated video will appear here
Create 4–30-second Seedance 2.5 videos from text, a start image, or image, video, and audio references. Choose 480p or 720p for UGC and ecommerce ads.
Seedance 2.5 is an AI video generator available in Prizmad AI Studio. Its three workflows cover different starting points: Text-to-Video turns a written scene into a new clip, Image-to-Video animates a start image and can aim toward an optional end image, and Reference-to-Video uses uploaded images, videos, and audio as creative guidance. Every mode still requires a prompt, so you can direct the subject, action, scene, camera, lighting, pacing, and sound.
Reference-to-Video is the most controlled workflow. A generation can use up to 30 JPG, PNG, or WebP images at 30 MB each, 10 MP4 or MOV videos at 200 MB each, and 10 MP3 or WAV audio files at 15 MB each, with no more than 50 references in total. Video and audio references must each be 1.8–30.2 seconds. Address every upload in the prompt as @Image1, @Video1, or @Audio1. Reference audio must be paired with at least one image or video, while combined reference-video duration and combined reference-audio duration must each stay within 30.2 seconds.
For output planning, choose any whole-second duration from 4 to 30 seconds and either 480p or 720p. Text-to-Video and Reference-to-Video support Auto plus six fixed aspect ratios: 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16. Image-to-Video follows the uploaded start image's aspect ratio automatically. Use 480p for faster concept iteration and 720p when the extra detail matters for review or delivery.
Seedance 2.5 is one part of a connected creative workflow. Draft the angle with the free AI Ad Script Writer, create supporting stills in the AI Image Generator, add a presenter with the AI Avatar Generator, or finish a generated clip with Auto Captions. If you need a complete ad assembled from a product URL, start with the AI UGC ad generator instead.
Build vertical creator-style scenes with a clear hook, natural product interaction, and an explicit closing frame for TikTok, Reels, and Shorts creative testing.
Animate a product image into a reveal, demonstration, or lifestyle scene for a product page, paid-social concept, marketplace listing, or launch campaign.
Use separate product, character, location, motion, and audio references to give the model clearer roles and produce closer campaign variations from the same creative brief.
Generate a new scene from a prompt alone. Specify the subject, action, setting, camera movement, lighting, pacing, and desired sound instead of relying on a visual input.
Animate a JPG, PNG, or WebP start image up to 30 MB, with an optional end image for a directed transition. The output follows the start image's aspect ratio automatically.
Guide identity, products, locations, motion, or sound with up to 30 images, 10 videos, and 10 audio files. Call out each asset with @Image, @Video, and @Audio references in the prompt.
Choose Text-to-Video for a prompt-only scene, Image-to-Video to animate a start frame, or Reference-to-Video when identity, product, motion, style, or audio guidance matters.
Prepare the inputs for that mode. Use a clean, well-lit start image for animation; in reference mode, upload only assets that have a clear role in the final shot.
Write a director-style prompt: subject and action first, then environment, camera movement, lighting, pacing, and the final frame. Put spoken dialogue in double quotes.
In reference mode, connect instructions to the right files with @Image1, @Video1, and @Audio1. State what to copy from each reference rather than asking the model to infer every role.
Set a 4–30-second duration and choose 480p or 720p. Select the placement's aspect ratio before generation; Image-to-Video inherits the start image's shape automatically.
Decide whether to generate audio, review the live token estimate, and generate. Check product details, claims, continuity, dialogue, and placement-safe framing before publishing an ad.
480p is the practical choice for prompt and motion tests. Use 720p when you need more visible detail for stakeholder review or a delivery candidate. The selected frame size affects the token estimate.
More files do not automatically mean a better result. Give every image, video, or audio reference one clear job, cite it in the prompt, and remove conflicting examples before generating.
Token cost is calculated from the selected output and relevant inputs rather than presented as one flat price. The workspace shows the current estimate before you start the generation.
Seedance 2.5 is an AI video generator in Prizmad AI Studio with Text-to-Video, Image-to-Video, and Reference-to-Video modes. It creates 4–30-second clips at 480p or 720p and can generate synchronized audio.
Yes. Choose Text-to-Video and describe the subject, action, environment, camera movement, lighting, pacing, and sound. No image, video, or audio upload is required for that mode.
Upload a JPG, PNG, or WebP start image up to 30 MB, write the motion and camera direction, and optionally add an end image for a controlled transition. Image-to-Video follows the start image's aspect ratio automatically.
Published 2026-08-07 · Last updated 2026-08-07
Plan a longer single-pass sequence with timed beats such as hook, problem, demonstration, payoff, and end frame, then evaluate continuity before committing to a final media edit.
Generate native 9:16, 1:1, 16:9, 4:3, 3:4, or 21:9 concepts so each social, feed, landing-page, or widescreen placement starts with intentional framing.
Reference a movement clip or audio cue to explore camera rhythm, action timing, ambience, and effects before combining the strongest output with captions or a separate voiceover.
Choose any whole-second duration from 4 to 30 seconds. Short clips suit hook testing; longer outputs leave room for connected product beats or a compact ad narrative in one generation.
Use 480p to iterate more quickly or 720p for higher-detail output. Vertical, landscape, square, portrait, classic, and cinematic aspect ratios cover common ad placements.
Generate ambience, effects, and dialogue audio with the video at no additional token cost, or switch audio off when you plan to add a separate voiceover or music track.
Reference-to-Video accepts up to 30 images at 30 MB each, 10 MP4 or MOV videos at 200 MB each, and 10 MP3 or WAV audio files at 15 MB each, with 50 references maximum in total. Video and audio references must each be 1.8–30.2 seconds. Cite them as @Image1, @Video1, and @Audio1. Audio references require at least one image or video.
A single Prizmad generation can be any whole-second length from 4 to 30 seconds. This tool does not expose a multi-minute extension workflow, so plan on a maximum of 30 seconds per generated output.
Choose 480p or 720p. Text-to-Video and Reference-to-Video offer Auto, 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16. Image-to-Video uses the uploaded start image's aspect ratio automatically.
Yes. The Generate Audio switch can add synchronized ambience, effects, and dialogue audio at no additional token cost. Turn it off if you plan to add a separate AI voiceover or background music.
It can be used to create UGC-style concepts, product reveals, demonstrations, lifestyle scenes, and placement-specific ad variants. Always review generated product details, spoken or visual claims, continuity, and platform compliance before publishing.
There is no single flat token price for every configuration. The workspace calculates the estimate from the selected output and relevant inputs, then shows the exact token amount before generation so you can adjust duration or resolution first.
Seedance 2.5 supports 4–30-second outputs and multimodal reference inputs: images, videos, and audio, up to 50 references in total. Seedance 2.0 supports 5-, 10-, or 15-second clips and up to 9 image references. Both offer 480p or 720p output and optional generated audio.
Reference-to-Video can use uploaded video clips as creative guidance, but this Prizmad tool does not expose a standalone timeline or region editor for modifying an existing video in place.