Image & Video Generation

The gateway exposes two media endpoints alongside chat: /v1/images/generations for images (synchronous) and /v1/videos for videos (asynchronous jobs). Both are billed on the provider's exact reported usage — no markup per call.

Authentication

Same as every other endpoint — send your Mume API key as a bearer token. See Authentication for how to create one.

Bash

Authorization: Bearer $MUME_API_KEY

Images

POST https://mume.ai/api/v1/images/generations returns the finished image in the response. The shape is OpenAI-compatible, so the OpenAI SDKs work unchanged.

cURL

Bash

curl https://mume.ai/api/v1/images/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $MUME_API_KEY" \
  -d '{
    "model": "google/gemini-2.5-flash-image",
    "prompt": "A lone lighthouse on a basalt cliff at dusk, volumetric fog, 35mm",
    "aspect_ratio": "16:9"
  }'

Python

import openai

client = openai.OpenAI(
    api_key="your-api-key",
    base_url="https://mume.ai/api/v1",
)

result = client.images.generate(
    model="google/gemini-2.5-flash-image",
    prompt="A lone lighthouse on a basalt cliff at dusk, volumetric fog, 35mm",
    extra_body={"aspect_ratio": "16:9"},
)

print(result.data[0].url)

Response

JSON

{
  "created": 1753401600,
  "id": "gen_01H...",
  "data": [{ "url": "https://firebasestorage.googleapis.com/..." }],
  "usage": { "cost": 0.078 }
}

Generated images are stored for you and returned as a hosted URL. Pass "response_format": "b64_json" to get raw base64 back instead.

Parameters

Parameter	Type	Description
model	string	Required. An image model slug, e.g. `google/gemini-2.5-flash-image`.
prompt	string	Required. What to generate.
aspect_ratio	string	`1:1`, `16:9`, `9:16`, `4:3`, `3:4`.
input_references	array	Reference images for editing / style transfer. Support and the maximum count vary by model.
response_format	string	`url` (default) or `b64_json`.

Editing with a reference image

Pass one or more images in input_references to edit or restyle them. URLs must be publicly reachable.

JSON

{
  "model": "google/gemini-2.5-flash-image",
  "prompt": "Make this scene look like a watercolour painting",
  "input_references": [
    {
      "type": "image_url",
      "image_url": { "url": "https://example.com/photo.jpg" }
    }
  ]
}

Videos

Video generation takes minutes, so it runs as a job. POST /v1/videos returns 202 with a job id immediately; poll GET /v1/videos/{id} until the status is completed or failed.

Submit a job

Bash

curl https://mume.ai/api/v1/videos \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $MUME_API_KEY" \
  -d '{
    "model": "google/veo-3.1",
    "prompt": "Slow dolly through a neon-lit ramen stall in the rain",
    "aspect_ratio": "16:9"
  }'

JSON

{
  "id": "vid_01H...",
  "status": "pending",
  "model": "google/veo-3.1"
}

Poll for the result

Bash

curl https://mume.ai/api/v1/videos/vid_01H... \
  -H "Authorization: Bearer $MUME_API_KEY"

JSON

{
  "id": "vid_01H...",
  "status": "completed",
  "model": "google/veo-3.1",
  "url": "https://firebasestorage.googleapis.com/..."
}

Status moves through pending → in_progress → completed | failed. On failure the response carries an error string. Credits are deducted once, when the job completes.

Image-to-video

Send a starting frame in frame_images to animate an existing image. frame_type is either first_frame or last_frame, and the URL must be publicly reachable.

JSON

{
  "model": "google/veo-3.1",
  "prompt": "The camera pushes in as the neon sign flickers",
  "frame_images": [
    {
      "type": "image_url",
      "image_url": { "url": "https://example.com/first-frame.jpg" },
      "frame_type": "first_frame"
    }
  ]
}

Polling example

Python

import time, requests

BASE = "https://mume.ai/api/v1"
headers = {"Authorization": f"Bearer {MUME_API_KEY}"}

job = requests.post(
    f"{BASE}/videos",
    headers=headers,
    json={
        "model": "google/veo-3.1",
        "prompt": "Slow dolly through a neon-lit ramen stall in the rain",
        "aspect_ratio": "16:9",
    },
).json()

while True:
    status = requests.get(f"{BASE}/videos/{job['id']}", headers=headers).json()
    if status["status"] in ("completed", "failed"):
        break
    time.sleep(8)

print(status.get("url") or status.get("error"))

Models & pricing

Image and video models use the same provider/model format as chat. Current families include Nano Banana (Gemini Image), GPT Image, FLUX.2, Seedream and Recraft for images, and Veo, Sora, Kling, Hailuo, Seedance and Wan for video. The Media Studio lists everything available to your account with live per-model pricing.

Billing is pass-through on the provider's reported usage: images are charged per image (or per megapixel/token, depending on the model) and videos per second of output. Some higher-cost models are limited to Plus accounts — see Pricing.

Errors

Media endpoints use the same error envelope as the rest of the gateway. A few you may hit specifically here:

Status	Meaning
402	Not enough credits for this generation.
403	The model is Plus-exclusive, or not served by the gateway.
404	Unknown video job id (or it belongs to another account).
502	The provider returned no media — safe to retry.

See Error Handling for the full list and retry guidance.

Image & Video Generation

Authentication

Same as every other endpoint — send your Mume API key as a bearer token. See Authentication for how to create one.

Bash

Authorization: Bearer $MUME_API_KEY

Images

POST https://mume.ai/api/v1/images/generations returns the finished image in the response. The shape is OpenAI-compatible, so the OpenAI SDKs work unchanged.

cURL

Bash

curl https://mume.ai/api/v1/images/generations \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $MUME_API_KEY" \
  -d '{
    "model": "google/gemini-2.5-flash-image",
    "prompt": "A lone lighthouse on a basalt cliff at dusk, volumetric fog, 35mm",
    "aspect_ratio": "16:9"
  }'

Python

import openai

client = openai.OpenAI(
    api_key="your-api-key",
    base_url="https://mume.ai/api/v1",
)

result = client.images.generate(
    model="google/gemini-2.5-flash-image",
    prompt="A lone lighthouse on a basalt cliff at dusk, volumetric fog, 35mm",
    extra_body={"aspect_ratio": "16:9"},
)

print(result.data[0].url)

Response

JSON

{
  "created": 1753401600,
  "id": "gen_01H...",
  "data": [{ "url": "https://firebasestorage.googleapis.com/..." }],
  "usage": { "cost": 0.078 }
}

Generated images are stored for you and returned as a hosted URL. Pass "response_format": "b64_json" to get raw base64 back instead.

Parameters

Parameter	Type	Description
model	string	Required. An image model slug, e.g. `google/gemini-2.5-flash-image`.
prompt	string	Required. What to generate.
aspect_ratio	string	`1:1`, `16:9`, `9:16`, `4:3`, `3:4`.
input_references	array	Reference images for editing / style transfer. Support and the maximum count vary by model.
response_format	string	`url` (default) or `b64_json`.

Editing with a reference image

Pass one or more images in input_references to edit or restyle them. URLs must be publicly reachable.

JSON

{
  "model": "google/gemini-2.5-flash-image",
  "prompt": "Make this scene look like a watercolour painting",
  "input_references": [
    {
      "type": "image_url",
      "image_url": { "url": "https://example.com/photo.jpg" }
    }
  ]
}

Videos

Video generation takes minutes, so it runs as a job. POST /v1/videos returns 202 with a job id immediately; poll GET /v1/videos/{id} until the status is completed or failed.

Submit a job

Bash

curl https://mume.ai/api/v1/videos \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $MUME_API_KEY" \
  -d '{
    "model": "google/veo-3.1",
    "prompt": "Slow dolly through a neon-lit ramen stall in the rain",
    "aspect_ratio": "16:9"
  }'

JSON

{
  "id": "vid_01H...",
  "status": "pending",
  "model": "google/veo-3.1"
}

Poll for the result

Bash

curl https://mume.ai/api/v1/videos/vid_01H... \
  -H "Authorization: Bearer $MUME_API_KEY"

JSON

{
  "id": "vid_01H...",
  "status": "completed",
  "model": "google/veo-3.1",
  "url": "https://firebasestorage.googleapis.com/..."
}

Status moves through pending → in_progress → completed | failed. On failure the response carries an error string. Credits are deducted once, when the job completes.

Image-to-video

Send a starting frame in frame_images to animate an existing image. frame_type is either first_frame or last_frame, and the URL must be publicly reachable.

JSON

{
  "model": "google/veo-3.1",
  "prompt": "The camera pushes in as the neon sign flickers",
  "frame_images": [
    {
      "type": "image_url",
      "image_url": { "url": "https://example.com/first-frame.jpg" },
      "frame_type": "first_frame"
    }
  ]
}

Polling example

Python

import time, requests

BASE = "https://mume.ai/api/v1"
headers = {"Authorization": f"Bearer {MUME_API_KEY}"}

job = requests.post(
    f"{BASE}/videos",
    headers=headers,
    json={
        "model": "google/veo-3.1",
        "prompt": "Slow dolly through a neon-lit ramen stall in the rain",
        "aspect_ratio": "16:9",
    },
).json()

while True:
    status = requests.get(f"{BASE}/videos/{job['id']}", headers=headers).json()
    if status["status"] in ("completed", "failed"):
        break
    time.sleep(8)

print(status.get("url") or status.get("error"))

Models & pricing

Errors

Media endpoints use the same error envelope as the rest of the gateway. A few you may hit specifically here:

Status	Meaning
402	Not enough credits for this generation.
403	The model is Plus-exclusive, or not served by the gateway.
404	Unknown video job id (or it belongs to another account).
502	The provider returned no media — safe to retry.

See Error Handling for the full list and retry guidance.