browser2api
Turn browser UIs into programmatic image and video generation APIs.
Automates web UIs via Playwright + Chrome CDP -- no API keys or reverse engineering needed.
Supported Platforms
| Jimeng (Ji Meng AI) |
Image: Seedream 5.0 -- up to 4 images, 2K/4K |
| Google Flow |
Image: Imagen 4, Nano Banana 2 -- up to 4 images |
Demo
Jimeng -- Image Generation with Seedream 5.0 * "A Shiba Inu eating ramen in a ramen shop, anime style"
jimeng_image_gen.mp4
Google Flow -- Image Generation with Nano Banana 2 * "Shiba eating ramen, anime style"
flow_image_gen.mp4
Google Flow -- Video Generation with Veo 3.1 Fast * "A cat walking through a garden"
flow_video_gen.mp4
Setup
python3 -m venv .venv
source .venv/bin/activate
pip install -e .
playwright install chromium
source .venv/bin/activate
pip install -e .
playwright install chromium
Requires Python 3.11+ and Google Chrome installed locally.
Quick Start
Jimeng -- Image Generation
python examples/generate.py "A cat in an astronaut suit standing on the moon"
python examples/generate.py "prompt" --model jimeng-5.0 --ratio 1:1 --resolution "Chao Qing 4K"
python examples/generate.py "prompt" --model jimeng-5.0 --ratio 1:1 --resolution "Chao Qing 4K"
Models: jimeng-3.0 jimeng-3.1 jimeng-4.0 jimeng-4.1 jimeng-4.5 jimeng-4.6 jimeng-5.0
Jimeng -- Video Generation
python examples/generate_video.py "A cat walking through a garden"
python examples/generate_video.py "City night skyline" --ratio 16:9 --duration 10s --model video-3.0-fast
python examples/generate_video.py "City night skyline" --ratio 16:9 --duration 10s --model video-3.0-fast
Models: seedance-2.0-fast seedance-2.0 video-3.5-pro video-3.0-pro video-3.0-fast video-3.0
Google Flow -- Image Generation
python examples/generate_flow.py "A sunset over mountains, digital art"
python examples/generate_flow.py "prompt" --model nano-banana-2 --orientation portrait --count 4
python examples/generate_flow.py "prompt" --model nano-banana-2 --orientation portrait --count 4
Models: nano-banana-pro nano-banana-2 imagen-4
Google Flow -- Video Generation
python examples/generate_flow_video.py "A cat walking through a garden"
python examples/generate_flow_video.py "prompt" --model veo-3.1-quality --orientation portrait --count 1
python examples/generate_flow_video.py "prompt" --model veo-3.1-quality --orientation portrait --count 1
Models: veo-3.1-fast veo-3.1-quality veo-2-fast veo-2-quality
How It Works
- Launch -- Opens a real Chrome browser via CDP (not Playwright's bundled Chromium) for anti-detection
- Login -- Checks for a valid session. If not logged in, opens a headed browser for manual login. Session persists in
~/.browser2api/browser_data/ - Generate -- Navigates to the generation page, configures model/settings via UI automation, fills the prompt, clicks submit
- Capture -- Intercepts network responses and polls the DOM for generated content
- Download -- Downloads results locally
Programmatic Usage
import asyncio
from browser2api import BrowserManager, Platform
from browser2api.platforms.jimeng import JimengClient, JimengModel, JimengRatio
async def main():
bm = BrowserManager()
context, page = await bm.launch_for_login(Platform.JIMENG)
client = JimengClient(
page, context,
output_dir="./output",
model=JimengModel.JIMENG_5_0,
ratio=JimengRatio.RATIO_1_1,
)
result = await client.generate_images("Yi Zhi Mao ", count=4, timeout_seconds=120)
print(f"Status: {result.status.value}")
for img in result.images:
print(f" {img.local_path} ({img.width}x{img.height})")
await bm.close()
asyncio.run(main())
from browser2api import BrowserManager, Platform
from browser2api.platforms.jimeng import JimengClient, JimengModel, JimengRatio
async def main():
bm = BrowserManager()
context, page = await bm.launch_for_login(Platform.JIMENG)
client = JimengClient(
page, context,
output_dir="./output",
model=JimengModel.JIMENG_5_0,
ratio=JimengRatio.RATIO_1_1,
)
result = await client.generate_images("Yi Zhi Mao ", count=4, timeout_seconds=120)
print(f"Status: {result.status.value}")
for img in result.images:
print(f" {img.local_path} ({img.width}x{img.height})")
await bm.close()
asyncio.run(main())
from browser2api.platforms.flow import FlowVideoClient, FlowVideoModel, FlowOrientation
async def generate_flow_video():
bm = BrowserManager()
context, page = await bm.launch_for_login(Platform.FLOW)
client = FlowVideoClient(
page, context,
output_dir="./output",
model=FlowVideoModel.VEO_3_1_FAST,
orientation=FlowOrientation.LANDSCAPE,
)
result = await client.generate_video("A cat walking through a garden", timeout_seconds=300)
print(f"Video: {result.video.local_path} ({result.video.size_bytes:,} bytes)")
await bm.close()
async def generate_flow_video():
bm = BrowserManager()
context, page = await bm.launch_for_login(Platform.FLOW)
client = FlowVideoClient(
page, context,
output_dir="./output",
model=FlowVideoModel.VEO_3_1_FAST,
orientation=FlowOrientation.LANDSCAPE,
)
result = await client.generate_video("A cat walking through a garden", timeout_seconds=300)
print(f"Video: {result.video.local_path} ({result.video.size_bytes:,} bytes)")
await bm.close()
Notes
Jimeng (Ji Meng AI)
- Member subscription recommended -- Required for video generation and higher daily image quotas.
- Login -- First run opens a headed Chrome window for manual login. Session is saved and reused.
- 4K resolution -- Requires membership. Free accounts are limited to 2K.
- Video generation -- 30-120s depending on model/duration.
Google Flow
- Google account required -- Must be logged in via the headed browser.
- Regional availability -- May require a VPN in some regions.
- Video generation -- 60-120+s depending on model.
Project Structure
src/browser2api/
+-- config.py # Shared constants (DATA_DIR)
+-- types.py # Platform, GenerationStatus, GeneratedImage, GeneratedVideo, etc.
+-- base.py # Abstract base classes
+-- browser.py # BrowserManager, ChromeLauncher, CDP connection
+-- platforms/
+-- jimeng/
| +-- client.py # JimengBaseClient, JimengClient, JimengVideoClient
| +-- enums.py # Models, ratios, resolutions, durations
| +-- selectors.py # CSS selectors and URL constants
| +-- login.py # Cookie-based login handler
+-- flow/
+-- client.py # FlowBaseClient, FlowClient, FlowVideoClient
+-- enums.py # Models, orientations, video models
+-- selectors.py # CSS selectors and URL constants
+-- login.py # Login handler
Disclaimer
This project automates browser interactions with third-party platforms. Use at your own risk.
- Rate limits -- Avoid aggressive usage. Respect platform rate limits and credit quotas.
- Session data -- Stored locally in
~/.browser2api/. Keep this directory secure. - For personal/research use only -- Do not use for commercial purposes or at scale without permission from the respective platforms.