Bernini-R

Generate videos from prompts or references, and edit existing videos with natural-language instructions.

Choose whether to generate a new video or edit an existing source video.

Describe the video to create or the exact edit to apply. Maximum 5000 characters.

Longer clips use more tokens. Actual output is approximately 5 or 7.5 seconds.

Token cost updates automatically because pricing scales by output long edge.

Sign in to spend tokens

Prompt is required

Your generated video will appear here

Overview

Bernini-R — Video Generation and Editing AI

Bernini-R is a video model that both generates new clips and edits existing footage from natural-language instructions — four modes in one workspace. Prizmad ships it as a built-in tool: a default 5-second clip at 848px costs 11 tokens, with the full range running 6–31 tokens depending on length and resolution.

Background

What is Bernini-R?

Bernini-R is a video AI model with an unusual scope: it doesn't just generate video from a prompt, it also edits video you already have. Its four modes — Text to Video, Reference to Video, Edit Video, and Reference Edit Video — cover the whole loop from creating a clip out of nothing to rewriting an existing shot with a plain-English instruction like turning the scene into a snowy winter setting while keeping the subject, framing, and motion intact.

The reference modes accept 1–5 images to guide the output — a product photo, a material swatch, a background, a style frame, or even weather reference. Edit modes take an uploaded MP4, MOV, or WebM source video up to 90 MB and apply your described change to it. Prompts can run up to 5,000 characters, and advanced controls include a seed for repeatable generations, an optional prompt-expansion pass that rewrites short prompts into richer video directions, and a speed setting that trades a little detail for faster turnaround.

On Prizmad, Bernini-R uses the same token balance as Seedance 2.0, Luma Ray 3.2, , , voiceover, and music. Pick a mode, write your prompt or edit instruction, choose length and resolution, and the finished MP4 lands in your asset library — no separate model account, no API keys, one token balance.

ChatGPT Image tools
AI avatars
Use cases

What you can create with Bernini-R

Seasonal ad refreshes

Take a performing ad clip and edit it into a winter, summer, or holiday version — same subject, same framing, new environment — instead of reshooting for every season.

Product scenes from reference photos

Reference to Video builds a continuous product scene around your actual product photos, so the packaging and details in the generated ad match what customers receive.

Cinematic product reveals

Text to Video handles premium reveal shots — slow push-ins, realistic lighting, polished commercial detail — for hero placements and landing-page headers.

Background and set swaps

Reference Edit Video combines an uploaded clip with reference images, letting you move a filmed subject into a new setting guided by a real background or style frame.

Vertical social variants

Generate 9:16 versions of a concept directly rather than cropping landscape footage, keeping the subject composed for TikTok, Reels, and Shorts placements.

Capabilities

Key capabilities

Four modes, one workspace

Text to Video and Reference to Video generate new clips; Edit Video and Reference Edit Video rewrite footage you upload. Switch modes at the top of the workspace without changing tools.

Natural-language video editing

Describe the exact change — swap the season, restyle the environment, alter the mood — and Bernini-R applies it to your source video while you specify what to keep, like subject, framing, and motion.

Up to 5 reference images

Guide generations or edits with 1–5 reference images for subject, product, material, background, style, or weather — useful for keeping a real product accurate in a generated scene.

Three resolution tiers

Choose a 576px long edge for speed, 848px for balance, or 1280px for the highest detail. The token cost in the workspace updates automatically because pricing scales with the output resolution.

5 or 7.5-second outputs

Pick a 5-second or 7.5-second clip in 16:9, 9:16, or 1:1 for generation modes — enough range to cover feed ads, vertical placements, and square posts.

Repeatable and tunable

Set an integer seed to reproduce a generation, enable prompt expansion to enrich short prompts automatically, or switch to Regular speed when turnaround matters more than maximum detail.

Workflow

How to use Bernini-R on Prizmad

1

Open the Bernini-R tool in Prizmad's AI Studio and pick a mode: Text to Video or Reference to Video to generate, Edit Video or Reference Edit Video to modify existing footage.

2

Write your prompt — up to 5,000 characters describing the scene to create or the exact edit to apply. Be specific about what should change and what should stay the same.

3

For edit modes, upload the source video (MP4, MOV, or WebM up to 90 MB). For reference modes, attach 1–5 images to guide subject, product, style, or environment.

4

Choose the output length (5 or 7.5 seconds), resolution (576, 848, or 1280px long edge), and — for generation modes — the aspect ratio (16:9, 9:16, or 1:1). The token cost updates as you change settings.

5

Optionally open advanced settings for a seed, prompt expansion, or the faster Regular speed mode.

6

Click Generate. Bernini-R typically returns the clip in about 3–8 minutes, and the MP4 lands in your asset library.

Pricing

How Bernini-R fits Prizmad pricing

Priced by length and resolution

The default configuration — 5 seconds at an 848px long edge — costs 11 tokens. Dropping to 576px cuts the cost and stepping up to 1280px or 7.5 seconds raises it, with the full range running 6 to 31 tokens. The workspace shows the exact cost before you generate.

Available for tokens

Bernini-R is unlocked on the same token balance you use for every other video and image tool — one token balance, no separate model access to buy.

Top up, don't upgrade

If tokens run out mid-project, a one-off top-up refills your balance directly. Generated and edited videos are yours with full commercial rights.

FAQ

Frequently asked questions

Is Bernini-R free to use?

It's available to any account with enough tokens rather than offered as a free standalone tool. Each generation draws tokens from your token balance — the default 5-second, 848px clip costs 11 tokens, and settings shift that between 6 and 31 tokens.

How much does a Bernini-R video cost?

Cost scales with output length and resolution. The default 5 seconds at 848px is 11 tokens; the cheapest configuration is 6 tokens and the most expensive (longer clip at 1280px) is 31 tokens. The workspace recalculates the price automatically as you change settings.

Can Bernini-R edit a video I already have?

Yes — that's its distinguishing feature. Edit Video and Reference Edit Video modes take an uploaded MP4, MOV, or WebM up to 90 MB and apply a natural-language instruction, such as changing the setting or weather while keeping the subject and camera framing intact.

What's the difference between Reference to Video and Edit Video?

Reference to Video generates a brand-new clip guided by 1–5 reference images (product, style, background, material). Edit Video modifies an existing uploaded video according to your prompt. Reference Edit Video combines both: it edits your uploaded footage while using reference images to guide the change.

What resolutions and durations does it support?

Outputs are approximately 5 or 7.5 seconds at a 576px, 848px, or 1280px long edge, in 16:9, 9:16, or 1:1 for generation modes. Higher resolution and longer duration both increase the token cost.

How does Bernini-R compare to Seedance 2.0 or Luma Ray 3.2 on Prizmad?

Seedance 2.0 and Luma Ray 3.2 are generation-focused — text, image, and reference to video. Bernini-R is the one to reach for when you need to edit existing footage with an instruction, or combine an uploaded video with reference images. All three run on the same token balance, so you can test them side by side without extra accounts.

Can I use Bernini-R videos commercially?

Yes. Videos you generate or edit on Prizmad carry full commercial rights — run them in paid ads, on product pages, and across social channels with no extra licensing.

How long does a generation take?

The workspace estimates roughly 3–8 minutes per job depending on length, resolution, and mode. Jobs run asynchronously, so you can queue more work while one renders. The Regular speed setting returns results faster at a slight cost to detail.

Related tools

More tools like this

LTX-2.3 ReframeConvert videos to vertical, landscape, square, or social formats without destructive cropping. Newly exposed areas are generated to match the original scene.Gemini Omni FlashCreate and edit realistic videos from text, images, or references with synchronized audio.Grok Imagine Video 1.5Turn a reference image into a polished video with smooth camera motion and coherent scenes.VEED Lip Sync 2.0Replace a video's audio and automatically match the speaker's lips, emotion, delivery, and timing.Bria RemovalBackground removalPixelcut Video Background RemovalRemove a video background frame by frame while keeping motion, edges, and fine details consistent.

Published 2026-07-16 · Last updated 2026-07-16