Together AI

Together AI (/media/togetherai) is an AI-hub studio that reuses your existing Together AI LLM key to generate images, videos, and speech through one inference cloud. Its tagline is “The AI-native cloud for open-source models.”

Where to configure

This studio uses universal-credential reuse: it reads the Together AI key you already configured under Settings → AI. No separate Settings → Media credential is needed, but the provider only appears as configured when that key exists. See Settings.

Tabs / operations

Tab	Operation	Output	Key fields
Text → Image	Image	Still image	Model, prompt, width, height, steps, number of images, seed.
Text → Video	Video	Video clip	Model (typed id), prompt, negative prompt, aspect ratio, duration, generate-audio toggle, seed.
Image → Video	Video	Video clip	Model (typed id), source image, prompt, aspect ratio, duration.
Text → Speech	Audio	Voiceover	Model, text, voice, response format (MP3/WAV).

The image and TTS tabs use Together's OpenAI-compatible endpoints. The video tabs use Together's native /v1/videos async API.

Generation flow

Image and TTS generations are synchronous and appear immediately. Video generations are asynchronous and finish through the Render Queue. All completed artifacts are saved to /files and can be opened in the Designer or posted. For the shared flow, see Media Studios overview.

Caveats

• Video has no webhook; completion relies on the media-jobs-poll cron.
• The model dropdown is populated live from Together's /v1/models catalog for image models, with a curated fallback list for Text → Speech. Because Together does not reliably tag every modality, the combobox also lets you type a model id directly — the video tabs have no curated list, so you type the video model id there.
• The source image for Image → Video is resolved server-side to a provider-reachable URL before submission.

Related docs

• Media Studios overview
• Media Library
• Settings

---

Verified against v1.3.0