Fractalyze/kandinsky6-rtx5090-showcase

text-to-video model by Fractalyze

From the publisher

Model card & documentation

Source preview

Read the publisher’s intended use, setup instructions, evaluations and limitations. The original model card is the source of truth.

Kandinsky 6.0 Pro-distill (29B joint video+audio DiT, 56 GiB in BF16) generating 864x480, 121 frames at 24 fps with audio on a single RTX 5090 (32 GB) in a 60 GB host, with vLLM-Omni. Upstream cannot serve this checkpoint on that machine; this branch does it in 151 s, 36% faster than the BF16 reference configuration, at LPIPS 0.11 against that reference. A fast, lossy mode takes 105 s (Results). Every change is behind a switch, off by default. W1: 864x480, 121 frames, 10 PiFlow steps, guidance 1.0, audio on, one RTX 5090, batch 1, seed 42. Speed: median of a mirrored A B B A session. Quality: LPIPS mean / worst frame over nine prompts, against a compiled BF16 reference generated on the same code and host. The reference's wall time is from its own gate run, not the same session. W2 (Pro-5s, 50 steps, CFG 5.0): 1611 s…

Fractalyze/kandinsky6-rtx5090-showcase on Hugging Face A short preview, not the full document.

Read the full model card ↗ · Preview checked 2026-10-08T08:17:14.625Z

Inside the original model card — Document outline

Headings are captured from the source. Links open the publisher’s document, not a locally hosted copy.

Links from the model card

References supplied by the publisher, not independently verified endorsements. Check the destination before downloading files or entering credentials.

showcase/kandinsky6/README.md

Documentation belongs to its respective authors. Reported project/model license: mit. A listing is not a grant of reuse or training rights. Confirm the document’s own terms at the source.

Model overview

Fractalyze/kandinsky6-rtx5090-showcase is a text-to-video model repository published by Fractalyze on Hugging Face. A library was not reported.

This page summarizes Hub metadata. For intended use, training data, evaluation results and limitations, consult the original model card.

Model facts

Task: text-to-video
Library: Not reported
Recent downloads (30 days): 0
Cumulative likes: 0
Hugging Face trending score: 0
Reported safetensors parameters: Not reported
Architecture: Not reported
License: mit
Access: Not gated by Hugging Face
Created: 2026-10-08T06:14:46.000Z
Last modified: 2026-10-08T06:15:05.000Z

Compatibility and lineage

Language coverage was not reported.

Reported base models: kandinskylab/Kandinsky-6.0-Pro-distill-5s-Diffusers

Parameter count is not a RAM/VRAM requirement. Check precision, quantization, context length and runtime compatibility in the model card; no hardware or API-cost claim is inferred here.

Use and evaluate

Check license terms and access requirements first. Review the model card’s documented loading instructions, evaluations and safety limitations. Benchmark scores are not imported by this directory, and popularity is not an accuracy ranking.

No model weights are downloaded or executed by AltAPIs. Never enable remote model code without reviewing it.

Model card and usage instructions

Model files and configuration

Community discussion