Turn text into an image on Kanvora

Name the subject, action or state, environment, composition, light, material or style, and exclusions. Decide what success means before choosing a model. The model can interpret the brief, but it cannot know which unstated detail matters to you.

• Subject: one primary person, object, place, or abstract form.
• Action or state: what the subject is doing or how it is arranged.
• Environment: the location, background, time, or spatial context.
• Composition: camera distance, view, focal point, and empty-space needs.
• Light and style: observable visual treatment rather than a vague quality claim.
• Exclusions: extra subjects, watermarks, unowned logos, tiny text, or unsafe content.

The template is deliberately compact. Add only details you will inspect. A prompt that names ten competing subjects and five visual styles creates more ambiguity, not more control.

Subject: [one primary subject]
Action or state: [what is happening]
Environment: [place, background, time]
Composition: [view, framing, focal point, negative space]
Light and visual treatment: [observable lighting and style]
Must include: [two or three reviewable details]
Must avoid: extra subjects, watermarks, logos, tiny text, unsafe or deceptive content

Subject: one compact coastal observatory with a cobalt dome
Action or state: telescope doors open toward the evening sky
Environment: rocky headland at blue hour with a calm sea
Composition: wide 16:9 frame, observatory on the left third, open sky on the right
Light and visual treatment: crisp cut-paper forms with long precise shadows
Must include: one warm window, clear horizon, generous empty sky
Must avoid: people, logos, text, extra buildings, watermarks

Kanvora exposes four model paths behind one workflow. Seedream 5.0 Lite is fixed 2K at 3 credits. Seedream 4.5 is fixed 2K at 4 credits. Seedream 5.0 Pro is fixed 1K at 6 credits. Nano Banana 2 offers 1K at 4 credits or 2K at 5. All four use the same five aspect ratios.

Resolution affects output size, not truth, typography accuracy, identity fidelity, or copyright clearance. Review every image before use.

Explore every public setting before sign-in. Authentication is required only when Generate image would create a job and place a credit hold.

1. 1

   Write the smallest complete brief

   Enter 3 to 2,000 characters covering subject, setting, composition, treatment, and exclusions.

   ContinueEach important visual requirement can be checked in the final image.Stop or reviseThe prompt depends on a missing reference image, exact identity, or unsupported output type.
2. 2

   Choose model, size, and ratio

   Compare the displayed credit cost, select Seedream 5.0 Lite, Seedream 5.0 Pro, Seedream 4.5, or Nano Banana 2, then choose the destination frame.

   ContinueThe setting matches the review need and available credit budget.Stop or reviseDo not choose a larger output only to compensate for an unclear brief.
3. 3

   Sign in and submit once

   Generate image opens the sign-in gate for guests. After authentication, security checks run before the job and credit hold are created.

   ContinueOne queued job and one integer credit hold appear.Stop or reviseDo not create a duplicate while the request is active.
4. 4

   Follow the original queue

   QUEUED and RUNNING are active. UNKNOWN means Kanvora is reconciling an uncertain provider submission rather than retrying it blindly.

   ContinueThe job reaches a stored success, stored partial result, or definite terminal failure.Stop or reviseWait when status is UNKNOWN. A definite failure releases the hold.
5. 5

   Inspect and download the stored still

   Review required details, unsafe or misleading marks, text, and destination crop before download or publication.

   ContinueThe stored file downloads and every locked check passes.Stop or reviseRevise one failed field or leave generation when direct editing is required.

A generation can be visually coherent and still fail the task. The check is anchored to the brief, not the effort already spent.

Avoid restarting from scratch after every failure. Preserve passing parts of the brief and revise one content, composition, treatment, or exclusion field at a time.

Subject is wrong

Reduce the scene to one primary subject and name the few defining details that must survive.

Composition is crowded

Remove secondary objects, name the focal position, and reserve negative space for the destination.

Style is inconsistent

Choose one visual treatment and describe its observable materials, edges, light, and color behavior.

Text is inaccurate

Shorten or remove embedded text. Use a layout tool when exact words must be editable and correct.

Unowned mark appears

Do not publish it. Exclude logos and watermarks, regenerate, and inspect the next file again.

Queue status is UNKNOWN

Wait for reconciliation of the original request. Never use a second submission as a status check.

A one-time signup grant adds 20 credits. That covers up to six Seedream 5.0 Lite attempts with 2 credits left, five Seedream 4.5 or Nano Banana 2 1K attempts, or four Nano Banana 2 2K attempts. Mixed-model totals depend on the settings you actually submit.

Purchased credits do not expire. A failed or rejected creative review still used the provider attempt when the job completed, while a definite provider failure releases its hold according to the runtime flow.

Use Kanvora for still-image generation from text. Leave this workflow when the next step requires an uploaded source, direct pixel editing, exact layered typography, or motion.

Generate on Kanvora

A text prompt and one of five ratios are enough to create the still you need.

Try another live model

The brief is valid, but you want to compare Nano Banana 2 and the three Seedream text-to-image paths within a pre-set credit budget.

Wait for the current job

The original submission is QUEUED, RUNNING, or UNKNOWN and must settle before retry.

Use an image editor

The content passes, but exact typography, masking, compositing, or layer edits are required.

Use another product

The task requires image-to-image, a reference upload, video, 4K Nano Banana 2, or another unshipped model.

Keep the current proof

Every locked detail passes and the stored file downloads. More variation is not completion.

Final review checklist

• The subject, action, environment, and focal point match the brief.
• No unsafe, deceptive, unowned, or unintended mark appears.
• Any visible text has been checked character by character.
• The ratio fits the destination and the focal subject survives the crop.
• The result is stored and the authenticated download responds.

The checked-in Kanvora model registry is authoritative for live IDs, sizes, and integer credits. Google, ByteDance, and fal document the upstream model paths. Upstream documentation does not promise that a particular prompt will be accurate, safe to publish, or suitable for a specific destination.

• Google AI: Nano Banana 2 API image generation documentation.
• ByteDance: Seedream 5.0 Lite model page.
• fal: Seedream 5.0 Lite text-to-image endpoint and Seedream 4.5 text-to-image endpoint.
• Kanvora: Live generator, pricing, Acceptable Use Policy, Terms, and Privacy.

Last verified: August 20, 2026. Frozen PAP review due: November 20, 2026.