Media Studios

/media is Postmill's collection of generative media tools. It is tools only — there is no asset library inside it. Every tool either produces a new file or transcribes an existing one, and every finished asset lands in the Media Library at /files, which remains the single place you browse, organise, and reuse everything you've generated or uploaded.

Nav structure

The /media sidebar groups 46 tools into three sections:

Section	Contents
Platform	Designer (manual Konva canvas/timeline editor) and AI Designer (chat-driven canvas assistant).
Providers	38 generation studios, one per provider — text-to-image/video/audio models, AI avatar video, music, and more.
Content Pack	6 stock browsers — Stock Photos, Stock Videos, Vectors, Stickers, Stock Audio, Icons.

Provider studios that aren't configured yet stay visible but dimmed; clicking one opens its own landing screen with a Configure link into Settings → Media. Configure credentials once per provider — no environment-variable fallback for any paid provider.

The standard generation flow

Most provider studios in the Providers section are built on a shared Studio Kit, so they all work the same way:

1. Pick a source asset (when the tool needs one) — image-to-video, image-to-image, and similar operations open the Media Selector, which lets you choose a file already in /files (it validates the picked file matches the field's expected kind — image, video, or audio).
2. Fill in the form — prompt, model, and any provider-native parameters (resolution, duration, style, etc.). Required fields are marked; the Generate button stays disabled until they're filled.
3. Generate — the studio posts the job to the backend and the request appears in the Render Queue, a live panel docked to the right of the form.
4. Track progress — the Render Queue polls while any job is pending/processing and shows a status pill (Queued, Rendering, Ready, Failed). Image and audio/TTS generations from most providers complete synchronously and appear almost immediately; video generations typically run as background jobs.
5. Land in Files — a completed job's artifact is already saved into your media storage; the queue card shows an inline preview (image, video, or audio player).
6. Hand off — every completed image/video/audio card offers two buttons:
   • Edit in Designer — opens the artifact in the Designer (images land on the static canvas, audio/video land on the video timeline).
   • Post — opens the composer pre-filled with the artifact, ready to schedule.

Two exceptions

Not every tool in /media follows the generation flow above:

• Content Pack (stock) browsers search a free or premium stock catalog and save a chosen result straight into /files — there's no generation, no render queue, and no Designer/Post handoff step baked into the search itself (a saved file can still be opened in Designer afterward like any other file).
• Deepgram transcribes speech to text. It returns { text, words, segments } instead of a media file, so its render-queue card is a text card (Copy / To composer) with no preview and no Edit-in-Designer/Post buttons. Its one Designer handoff is different too: it can send a video source plus its computed captions into the Designer, which burns them in as a caption track without re-transcribing.

Bespoke studios

A few tools have enough of a unique interaction model that they keep their own bespoke UI instead of riding the generic Studio Kit form:

• Designer — the manual canvas/timeline editor itself, and the target of every "Edit in Designer" handoff.
• AI Designer — a conversational agent that drives the Designer for you.
• HeyGen — four structured tabs (Storyboard, Talking Photo, Voiceover, Translate) for AI avatar video, still landing in the same render queue / Files / Post pattern.
• Replicate — a 19-category workspace (image, video, audio, STT, inpainting, merge, memes) with its own model picker, cost badges, and specialised editors (mask painter, clip merge, meme layers).
• Deepgram — speech-to-text and caption export, described above.

All studios

Platform: Designer · AI Designer

Providers: Replicate · HeyGen · Deepgram · Reel.Farm · Genviral · Kling · Higgsfield · LTX Studio · Luma · MiniMax · Pika · Qwen · Together AI · Runway · Suno · Wan · SiliconFlow · Groq · OpenRouter · Fireworks AI · DeepInfra · xAI Grok · Vercel AI Gateway · Amazon Bedrock · Azure OpenAI · Google AI Studio · Google Vertex · Black Forest Labs · Stability AI · Recraft · Ideogram · Leonardo.ai · OpenAI · Sora · ElevenLabs · D-ID · Hedra · Tavus

Content Pack: Stock Photos · Stock Videos · Vectors · Stickers · Stock Audio · Icons

Most provider studios are configured under Settings → Media — see Settings for credential setup. Some studios reuse a key you may already hold elsewhere: the universal-credential studios (Qwen, Google AI) accept the same key configured at Settings → AI, and a few ride another provider's credential entirely (Sora uses your OpenAI key, Pika uses your fal key). Each studio's own page notes where its credential lives.

---

Verified against v1.0.0