Image generation

Image generation

Turn a text prompt into an image. Describe what you want, get a high-resolution PNG back, and find it waiting in your workspace’s Drive library — ready to reuse across every Destesi product.

You can work in the web app — a chat interface with a gallery of everything you’ve made — drive it from your own code over a small HTTP API, or ask for an image in Chat and have it generated inline.

Two engines

Image generation is not one model. GET /v1/engines reports what your workspace can use:

Engine	Id	What it is	Availability
Fast	default	The platform engine. Quick, and always there.	Always available
High quality (GPT)	openai	A higher-quality model that runs on your own OpenAI account, connected through Connect.	Shows as unavailable until you connect OpenAI

The engine selector in the web app is the same list. Omit engine on an API call and you get the fast engine.

Bring your own OpenAI account

The high-quality engine bills to your OpenAI account, not to Destesi. Connect OpenAI once per workspace in Connect and the engine unlocks; revoke it there and the engine goes back to unavailable.

Generate your first image

The fastest path to a result is the web app.

1. Sign in. Go to account.destesi.io and sign in. Your active workspace is the one you set up during account onboarding, and everything you generate is scoped to it. (Need to change it? Use the account menu in the top-right of any product.)

2. Open Image generation. Click the 9-dot launcher and choose Image generation (or go straight to imagegen.destesi.io). Your single sign-on session follows you in — no second login.

3. Describe the image. The app is a chat interface: type what you want and send it. Pick an engine from the selector first if you want the high-quality one. You’ll see live progress as the model works, then the finished image appears inline.

4. Browse the gallery. Everything you’ve generated in this workspace is in the gallery, so you can find an earlier image without hunting through the conversation.

5. Find it in Drive. Every image is also registered in your workspace’s Drive library under the ImageGen generated folder, so it’s there for any product to reuse.

Generate from your own code

Image generation also speaks a small HTTP API at https://api.imagegen.destesi.io. Authenticate either with the imagegen_session cookie your browser already holds, or with a Personal Access Token plus the workspace to run in:

Authorization: Bearer idn_pat_<your-token>
X-Destesi-Workspace: <your-workspace-slug>

The streaming endpoint is the one to reach for from a UI — it opens a single connection and pushes progress events until the image is ready:

curl -N https://api.imagegen.destesi.io/v1/generate-and-stream \
  -H 'Content-Type: application/json' \
  -H "Authorization: Bearer $DESTESI_PAT" \
  -H "X-Destesi-Workspace: $DESTESI_WORKSPACE" \
  -d '{"prompt": "a watercolor fox in a misty forest, soft light"}'

You’ll receive a Server-Sent Event stream:

event: queued
data: {"job_id":"a1b2c3d4e5f6a7b8"}

event: progress
data: {"job_id":"a1b2c3d4e5f6a7b8","step":12,"total":30}

event: done
data: {"job_id":"a1b2c3d4e5f6a7b8","image_url":"https://drive.destesi.io/...","drive_file_id":"file_..."}

The done event carries image_url — a ready-to-load URL for the rendered PNG — and drive_file_id, the file’s permanent home in Drive.

Prefer a single blocking request? POST /v1/generate waits for the result and returns it in one JSON response, up to a 120-second ceiling. See the API reference for every field, endpoint, and error code.

Tune the output

Pass engine, negative_prompt, width, height, steps, and guidance_scale to pick the model and control resolution and prompt adherence. width and height are a fixed set — 512, 576, 768, or 1024 — and steps is capped at 50. Defaults (1024×1024, 30 steps, guidance 7.5, fast engine) apply when you omit them.

Where it fits in the suite

Image generation is a peer product in the Destesi suite. It shares the same account, the same workspace, and the same library as everything else:

• One sign-in via Accounts, one workspace that owns your data.
• Every image is registered in Drive — one library across the suite, so Chat, Studio, and your own tools can reuse what you make.

Next steps

ConceptsThe mental model and how it fits the suite.

API referenceEndpoints, request fields, env vars.

TroubleshootingCommon failures and fixes.