Dezgo

Dezgo for browser-based image creation and editing

Dezgo provides AI image and video tools through a web interface, with separate modes for generation, image-to-image changes, controlled composition, inpainting, background removal and upscaling. It suits illustrators exploring concepts, creators making supporting visuals and developers who want image operations through an API without maintaining a local model installation.

The interface presents model-specific forms rather than one identical set of controls for every operation. That matters when choosing a workflow: a text prompt starts a new picture, a reference can guide structure, and an editing operation changes an existing image.

Generation controls and reference structure

The Flux text-to-image form includes prompt, image shape, guidance, sampling steps and seed controls. A seed is useful when documenting a generation setup, though changing the model or other settings can still change its result. Some options, including particular batch and LoRA controls, are marked for Power Mode.

ControlNet adds an uploaded control image and a control type such as Canny edges. Control strength and preprocessing determine how the reference structure contributes to the generated image. The documented form automatically resizes its input to a maximum of 512 by 512 pixels; that limit belongs to this mode and should not be treated as a universal limit across Dezgo.

Targeted edits and programmatic access

Dezgo's text-inpainting mode separates the description of the area to replace from the description of its replacement. For example, identifying a cushion and asking for a blue linen cushion serve different purposes. This mode also exposes negative prompts and sampling settings. Other operations cover image upscaling and background removal.

The official developer documentation lists endpoints for generation, ControlNet, image-to-image, inpainting and related image tools. Dezgo says its current API and web Power Mode use the same balance. Its older RapidAPI offering is described as a separate, deprecated service, so developers should follow the current documentation when starting an integration.

Example: develop a room illustration

Begin with a simple room sketch or edge reference you have permission to use. In ControlNet, select the matching control type and describe the desired materials and lighting. Compare a few settings while keeping a record of the model, seed and control strength. If the layout works but a cushion is wrong, use text inpainting to identify that area and describe the replacement.

Inspect corners, furniture joins and the edited boundary before upscaling. Save the result and its settings outside the service for later revisions. This workflow supports an illustration or mood board; it does not establish measurements, structural feasibility or the factual appearance of an unbuilt room.

Limitations that affect a request

Controls and output limits vary by model and mode. Higher guidance is not automatically better, and the ControlNet form warns that extreme values can introduce artifacts. Automatic masks can select the wrong area, while reference structure may constrain an unwanted detail. Review each model's license and the rights to uploaded material before using outputs in client work.

Free access and prepaid Power Mode

Dezgo's FAQ describes free use without registration, subject to rate, resolution and sampling limits. Some compute-intensive features require paid Power Mode. Power Mode uses a prepaid balance rather than a recurring subscription, with request cost varying by operation and settings. An account is accessed through an emailed login link. Check the displayed request price before running a paid job.