Kandinsky 6.0 Pro

Positioning Overview

Kandinsky 6.0 Pro is a text-to-video model aimed at developers who need to get a video with sound in a single pass. It takes a prompt as its only required input and produces synchronized audio alongside the generated visuals, outputting a clip of about 5 seconds per run, with support for landscape, portrait, and square aspect ratios, plus a random seed for reproducing results. It suits short-video asset prototyping, voiceover and lip-sync shot validation, and teams that iterate quickly on creative ideas using English or Russian prompts.

Core Highlights

• Visuals and audio generated in one pass — A prompt can describe the subject, action, camera, and sound at the same time; visuals and audio are produced in the same pipeline, eliminating the need for later voiceover alignment.
• Reproducible generation control — A seed parameter is provided; once fixed, it makes it easy to compare versions and run regressions on the same prompt.
• Coverage of multiple aspect ratios — Supports five output frames: 16:9, 9:16, 1:1, 4:3, and 3:4, allowing direct switching between landscape, portrait, and square finished videos.
• Separate prompt and negative control — guidance_scale adjusts how closely the prompt is followed, while negative_prompt separately declares content to avoid.

Capability Description

Prompt-to-video with sound: Enter a prompt describing both visuals and sound to get a short clip with synchronized audio. English and Russian prompts are natively supported, making it easy to control the content and mood of the finished video directly with natural language.

Aspect ratio switching: Choose between landscape, portrait, and square via aspect_ratio, so the same prompt can quickly produce versions suited to different placements.

Adherence and negative constraints: guidance_scale adjusts how strictly the text is followed, from 1 to 10, while negative_prompt excludes unwanted elements; used together, they narrow down trial and error.

Seed reproducibility: seed can fix randomness or, if left empty, generate randomly, making it suitable for locking results during finalization or running controlled experiments during tuning.

Applicable Scenarios

• Rapid production of social media short-video assets — Operators or use-case validators enter a prompt describing visuals and sound effects and directly get a roughly 5-second finished clip, used as a prototype for portrait or square placement assets.
• Shot concept validation with voiceover — Creative teams use prompts to describe character actions and sound, generating short shots with audio to demonstrate narrative pacing and mood to clients or teammates.
• Batch adaptation to multiple aspect ratio versions — Developers who need the same creative concept to cover both landscape and portrait channels fix the seed and prompt, switch aspect ratios to produce multiple asset versions, and reduce repeated adjustments.

Limitations and Selection Advice

A single output is a short clip of about 5 seconds; narrative long videos need to be generated in segments and then stitched together. Generating with audio means the prompt must cover both visual and sound descriptions, and under complex compositions, multiple rounds of adjusting guidance_scale and the negative prompt may be needed to converge. Currently only text input is supported; if you already have static assets and want them to drive dynamic effects, use the image-to-video route in the same series instead. For scenarios with higher requirements on finished video duration and resolution, evaluate whether to combine with subsequent processing such as super-resolution.