google/nano-banana/editNano Banana Edit refines images using natural language prompts and reference photos, supporting localized edits and background swaps. It strictly preserves facial identity, posture, and lighting harmony during modifications.
Get API Key
Use this photograph as the only source. Change only the middle tile: replace the celadon-green glaze with a deep glossy cobalt-blue glaze, and change its handwritten label from "CELADON" to "COBALT BLUE" in the same letter size, paper and handwriting. Keep the left copper-red tile and "COPPER RED" label, the right iron-black tile and "IRON BLACK" label, the plaster wall, lighting and camera angle exactly. No other tiles, no extra text, no logos, no watermarks.
Add only a sleeping orange tabby cat curled on the center thwart of this same boat, and a folded grey wool blanket tucked beside it. Keep the boat, oars, mist, water and camera angle exactly. The cat must have realistic weight and a contact shadow on the wood. No people, no extra objects, no text, no logos, no watermarks.
Use image 1 as the identity reference and image 2 as the destination. Place this same woman sitting cross-legged on the bamboo raft, cutting a sheet of red paper with small scissors, a few red scraps resting on the raft. Preserve her face, age, silver bun, indigo coat and ink-stained fingers from image 1. Preserve the river, limestone cliffs and raft from image 2. Natural overcast light, physically consistent contact with the raft, one woman only, no collage edges. No extra people, no text, no logos, no watermarks.
REST API
This flow submits one valid Nano Banana Edit request, captures its task_id, and reads image URLs after the asynchronous task ends.
Create an API key, keep it on the server, and send it as Authorization: Bearer VIDGO_API_KEY.
https://api.vidgo.ai/api/generate/submitSend the current model ID, prompt, and ordered image URLs to the unified submit endpoint. The example keeps only first-call fields.
curl --request POST \
--url "https://api.vidgo.ai/api/generate/submit" \
--header "Authorization: Bearer $VIDGO_API_KEY" \
--header "Content-Type: application/json" \
--data '{
"model": "google/nano-banana/edit",
"input": {
"prompt": "Use this photograph as the only source. Change only the middle tile: replace the celadon-green glaze with a deep glossy cobalt-blue glaze, and change its handwritten label from \"CELADON\" to \"COBALT BLUE\" in the same letter size, paper and handwriting. Keep the left copper-red tile and \"COPPER RED\" label, the right iron-black tile and \"IRON BLACK\" label, the plaster wall, lighting and camera angle exactly. No other tiles, no extra text, no logos, no watermarks.",
"size": "4:3",
"image_urls": [
"https://cdn.vidgo.ai/apis/models/google/nano-banana-2/edit/v1/01/input-01.png"
]
}
}'Read data.task_id and poll only while status is not_started or running. Stop on both finished and failed.
GET https://api.vidgo.ai/api/generate/status/{task_id}Poll every 2–5 seconds and back off on 429/5xx. Or provide top-level callback_url to receive a terminal-state callback.
not_startedrunningfinishedfailed{
"code": 200,
"data": {
"task_id": "3UBJ3ROCRTFMUSL3",
"status": "running",
"created_time": "2026-09-26T10:32:02"
}
}{
"code": 200,
"data": {
"task_id": "3UBJ3ROCRTFMUSL3",
"status": "finished",
"files": [
{
"file_type": "image",
"file_url": "https://cdn.vidgo.ai/apis/models/google/nano-banana-2/edit/v1/01/output.jpg"
}
],
"created_time": "2026-09-26T10:32:02",
"error_message": null,
"progress": 100
}
}Expand for the same payload end to end, including business-code, missing-task-ID, failed-state, unexpected-state, and network-error handling.
set -euo pipefail
: "${VIDGO_API_KEY:?Set VIDGO_API_KEY in your environment}"
REQUEST_BODY=$(cat <<'JSON'
{
"model": "google/nano-banana/edit",
"input": {
"prompt": "Use this photograph as the only source. Change only the middle tile: replace the celadon-green glaze with a deep glossy cobalt-blue glaze, and change its handwritten label from \"CELADON\" to \"COBALT BLUE\" in the same letter size, paper and handwriting. Keep the left copper-red tile and \"COPPER RED\" label, the right iron-black tile and \"IRON BLACK\" label, the plaster wall, lighting and camera angle exactly. No other tiles, no extra text, no logos, no watermarks.",
"size": "4:3",
"image_urls": [
"https://cdn.vidgo.ai/apis/models/google/nano-banana-2/edit/v1/01/input-01.png"
]
}
}
JSON
)
SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
--request POST \
--url "https://api.vidgo.ai/api/generate/submit" \
--header "Authorization: Bearer $VIDGO_API_KEY" \
--header "Content-Type: application/json" \
--data "$REQUEST_BODY")
TASK_ID=$(printf '%s' "$SUBMIT_RESPONSE" | jq -r '.data.task_id // .task_id // empty')
if [ -z "$TASK_ID" ]; then
printf 'Submit response did not include task_id:
%s
' "$SUBMIT_RESPONSE" >&2
exit 1
fi
while true; do
STATUS_RESPONSE=$(curl --silent --show-error --fail-with-body \
--url "https://api.vidgo.ai/api/generate/status/$TASK_ID" \
--header "Authorization: Bearer $VIDGO_API_KEY")
STATUS=$(printf '%s' "$STATUS_RESPONSE" | jq -r '.data.status // .status // empty')
case "$STATUS" in
finished)
printf '%s' "$STATUS_RESPONSE" | jq -r '(.data.files // .files // [])[]?.file_url'
break
;;
failed)
printf '%s' "$STATUS_RESPONSE" | jq -r '.data.error_message // .error_message // "Generation failed"' >&2
exit 1
;;
not_started|running)
sleep 2
;;
*)
printf 'Unexpected task status: %s
' "$STATUS" >&2
exit 1
;;
esac
doneThis table lists only public fields inside input. See Quick Start for top-level model and optional callback_url usage.
| Field | Type | Required | Default | Description |
|---|---|---|---|---|
| input.prompt | string | Yes | – | Describe the requested image changes using 1–5,000 characters, including at least one non-whitespace character. |
| input.image_urls | string[] | Yes | – | Reference images to edit. Provide at least one publicly accessible HTTP(S) image URL. |
| input.size | string | No | – | Output image aspect ratio. Select auto to let the model choose the framing; the API request then omits size. |
The submit response identifies the task. The status response supplies image files or failure details at a terminal state.
| Field | Type | Description |
|---|---|---|
| code | integer | Business response code; 0 or 200 indicates success. |
| message | string | Optional human-readable response message. |
| data.task_id | string | Unique task ID used for status queries after submission. |
| data.status | string | not_started, running, finished, or failed. |
| data.created_time | string | Task creation timestamp. |
| data.progress | integer | Progress from 0–100 in status responses. |
| data.files[] | array | Generated files in returned order when finished. |
| data.files[].file_url | string | Direct generated-image URL. |
| data.files[].file_type | string | Image results are identified as image. |
| data.files[].watermark_url | string | null | Optional watermarked-file URL when present. |
| data.error_message | string | null | Readable failure reason for failed tasks. |
Only the first two states may be polled again. finished and failed are terminal.
not_startedAccepted and queued; wait before checking again.
runningImage generation is in progress; continue polling.
finishedSuccessful terminal state; validate files and read image URLs.
failedFailure terminal state; read error_message and stop.
| Specification | Value | Details |
|---|---|---|
| Input mode | Prompt + images | At least 1 image; images participate in edit order from image_urls. |
| Prompt length | 1–5,000 | Counted in characters after trimming leading and trailing whitespace. |
| Output | Image | Returns image files asynchronously. |
| Aspect ratio | 10 ratios | Supports 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9. |
| Billing | Each generation costs 5 credits ($0.025). 1 credit = $0.005. | 1 credit = $0.005; the backend remains authoritative for final billing. |
Nano Banana Edit is a Google DeepMind model for precise image editing driven by natural language and visual references. It delivers breakthrough character and facial consistency, mask-free localized inpainting, background replacement, and multi-reference feature blending. Built on the Gemini 2.5 Flash unified vision-language architecture, it strictly preserves original subject poses, perspective geometry, and natural ambient lighting while executing targeted transformations. You can call it programmatically or try it from the playground above.
Leveraging Gemini's multimodal latent memory, the model anchors cranial proportions, facial geometry, and expressions from the source image. Reinforcing your prompt with instructions such as "preserve exact facial identity" ensures characters remain coherent across changing outfits and environments.
No. The model relies entirely on natural language semantic parsing to target modifications without manual masks or lasso selections. Simply state the desired change in your prompt (e.g., "change the red handbag to a brown leather backpack"), and the model automatically confines the edit to the designated item.
Yes. The model incorporates physical lighting re-estimation, recalculating contact shadows, rim reflections, and ambient color spill on the subject to match the light sources and color temperature of the new background.
Yes. You can supply multiple publicly accessible image URLs in the image_urls array to blend features across sources (e.g., transferring clothing style from a second image onto a character from the first), with no extra fee charged for additional reference pictures.
Yes. You can specify any of the 10 supported aspect ratios (1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, or 21:9) via the size parameter. If omitted or set to auto, the model automatically selects an optimal framing aligned with the original composition.
Image editing calls cost a flat rate of 5 credits ($0.025, where 1 credit = $0.005) per generation, with zero surcharges for multiple reference images. If a generation request encounters a system error or fails validation, all debited credits are refunded immediately.
