Skip to main content

Generative Models

Create images, audio/voice, and video from text (and images), served through the diffusers / generative backends.

Generative models are called through the same Inference API as everything else. They return artifacts (files) rather than text: the result carries an output_artifacts list of downloadable URLs, and you fetch them with client.endpoints.download_artifact(...). For image/audio inputs, upload a file first with client.endpoints.upload_input(...).

Snapshot — query the live catalog for the authoritative set
curl "https://api.colabhive.com/api/builder/v1/endpoints?visibility=public"

The active generative endpoints below are currently free (price_per_request = 0). Several are candidate (functional but not fully validated) — check readiness/lifecycle in the catalog before relying on one, and you can import more from HuggingFace.

Tiers​

🖼️ Image generation​

Text-to-image diffusion — FLUX.1-dev, Stable Diffusion XL, Stable Diffusion 3 Medium, Z-Image-Turbo. → Image Generation

🔊 Audio & voice​

Text-to-music / sound (MusicGen, AudioLDM2) and text-to-speech (Chatterbox, Kokoro, Bark, Parler-TTS). → Audio Generation

🎬 Video generation​

Text-to-video — Wan2.2 (TI2V-5B) and text-to-video-ms 1.7B. → Video Generation

🧊 3D generation​

Image-to-3D mesh generation. No active public 3D endpoint today — see the page for status and how to import one from HuggingFace. → 3D Generation

Common output shape​

Every generative endpoint returns output_artifacts:

{
  "output_artifacts": [
    {
      "filename": "image_0.png",
      "url": "https://storage.colabhive.com/...",
      "content_type": "image/png",
      "size_bytes": 1048576,
      "modality": "image"
    }
  ],
  "metadata": { "width": 1024, "height": 1024 }
}
from colabhive import ColabHive

client = ColabHive(api_key="hive_...", account_id="YOUR_ACCOUNT_ID")

result = client.endpoints.infer(
    endpoint_id="z-image-turbo",   # SDK resolves the name to a UUID
    input_data={"prompt": "A sunset over mountains, oil painting style"},
)
for art in result.get("result", {}).get("output_artifacts", []):
    client.endpoints.download_artifact(art["url"], f"./{art['filename']}")

Learn more​

Authors: José Luis Minich, Maximiliano Lucius.