<aside> 🛠

Real-Time Long Video Generation (Highest Quality Variant)

Most AI video tools either top out at a few seconds or need an expensive multi-GPU cluster to go longer — and quality drifts the moment a clip runs long. Helios Base is the flagship, highest-quality member of the Helios family: it generates long-form video (up to ~60 seconds) from a text description, a starting image, or an existing clip, and it does it in real time on a single GPU at 19.5 FPS. When output quality is the priority, this is the variant to reach for. Built by PKU-YuanGroup.

</aside>

What it does

Helios Base is the flagship, highest-quality variant of the Helios video generation model family, developed by PKU-YuanGroup. It generates long-form, visually rich video content from a text description, a reference image, or an existing video clip. While the distilled version of Helios prioritizes speed, Helios Base is the go-to choice when output quality is the top priority, producing the most detailed, coherent, and visually polished results in the family. It still runs on a single GPU and generates video in real time at 19.5 FPS, making it both high-quality and cost-effective compared to larger competing models.

Problem it solves

• Best-in-class output – Among the three Helios variants, Base delivers the highest visual fidelity, making it the right choice for premium creative and production use cases
• Cost-effective infrastructure – Runs on a single GPU rather than expensive multi-GPU server clusters, significantly reducing compute costs
• Flexible creative input – Works from a text description alone, a starting image, or an existing video clip, fitting a wide range of creative and production workflows
• Temporal coherence – Videos remain visually consistent and coherent throughout their full length, without the common "drift" or quality degradation seen in longer AI-generated clips
• Production-ready deployment – Designed to integrate into existing production pipelines and tools from day one

Input/Output

• Input:

  • Text to Video: Describe the scene you want in plain language and Helios generates the video

    

  • Image to Video: Provide a starting image and a description of the motion or scene

  • Video to Video: Provide an existing video clip and a description to transform or extend it

• Output: A fully generated MP4 video file

  • Up to ~60 seconds in length

  • High visual quality with smooth, coherent motion throughout

    

    55b41170-4bdf-4b41-aeb1-3743732256d1.mp4

• Parameters:

  • Resolution (W × H, ratio) (dropdown, default 640 × 384 (5:3)) — pick from 11 aspect-ratio presets:

    Landscape	Square	Portrait
    768 × 320 (12:5)	512 × 512 (1:1)	448 × 576 (7:9)
    768 × 384 (2:1)		512 × 768 (2:3)
    640 × 384 (5:3) (default)		384 × 640 (3:5)
    768 × 512 (3:2)		384 × 768 (1:2)
    576 × 448 (9:7)		320 × 768 (5:12)

  • Frame count (chunk multiples, default 99 frames (~4s @ 24fps)) — pick from a fixed list that maps to standard clip lengths at 24 fps:

    Frames	Duration @ 24 fps
    99 (default)	~4 s
    198	~8 s
    297	~12.5 s
    396	~16.5 s
    495	~21 s
    594	~25 s
    693	~29 s
    792	~33 s
    891	~37 s
    990	~41 s
    1,089	~45 s
    1,188	~49.5 s
    1,287	~54 s
    1,386	~58 s
    1,452	~60.5 s

    Frame counts must be chunk multiples — the model generates in 33-frame autoregressive chunks (99 = 3 chunks, 198 = 6 chunks, and so on), which is why the dropdown offers only these specific values.

  • Output FPS (dropdown, default 24) — choose 16 or 24 fps

    • 24 — standard cinematic frame rate
    • 16 — lighter alternative for shorter or more animation-style output

  • Advanced Options:

    • Diffusion steps (range 20 – 60, default 50) — how many denoising passes the model runs per chunk. More steps = higher quality, slower generation; fewer steps = faster, may lose some detail. This slider is the main quality/speed trade-off for the Base variant (the Distilled variant has this collapsed into a fixed schedule)
    • Random seed (-1 = random, default 1) — set a fixed integer to reproduce the same video across runs

Accuracy & Speed

• Speed & Scale

Helios Model Variants

Note: Image-to-Video and Video-to-Video modes may produce slightly less consistent results than Text-to-Video, as the model was primarily trained on text-based generation.