Higgsfield

Higgsfield (/media/higgsfield) is an AI-native creative suite for social content: Soul for text-to-image, DoP for cinematic image-to-video, and Speak for audio-driven talking video.

Where to configure

Configure Higgsfield under Settings → Media → Higgsfield. It uses a two-part credential — API Key ID and API Key Secret — rather than a single key. See Settings for credential setup.

Tabs

Tab	Operation	Key fields	Produces
Text → Image	Image	Prompt, size, quality, batch size, optional reference image, enhance prompt, seed	Still image(s)
Image → Video	Video	DoP model (Standard/Turbo/Lite), source image, prompt, enhance prompt, seed	MP4 video clip
Speak	Video	Portrait image, audio clip, prompt, quality, duration, seed	Talking-head video

Generation flow

All three tabs follow the standard Media Studios flow: jobs appear in the Render Queue and land in /files when complete. Images and videos can be edited in the Designer or posted.

Caveats

• Two-part auth. Credentials are sent as Authorization: Key <id>:<secret>. As a fallback, a single string in KEY_ID:KEY_SECRET form is accepted.
• Image is synchronous. Soul completes via bounded internal polling; batch_size: 4 returns multiple images, and each is saved as its own file.
• Video is asynchronous. DoP and Speak rely on the media-jobs-poll cron — Higgsfield provides no completion webhook.
• NSFW filter. A generation flagged as nsfw is reported as failed.
• Source image and audio media fields are resolved server-side to provider-reachable URLs.

---

Verified against v1.0.0