Update app.py
Browse files
app.py
CHANGED
|
@@ -10,21 +10,18 @@ import imageio
|
|
| 10 |
import torch
|
| 11 |
from custom_wan_pipeline import WanImageToVideoPipeline
|
| 12 |
|
|
|
|
| 13 |
app = FastAPI()
|
| 14 |
|
| 15 |
-
# ===== LOAD YOUR MODEL =====
|
| 16 |
print("Loading WAN I2V model...")
|
| 17 |
pipe = WanImageToVideoPipeline.from_pretrained(
|
| 18 |
-
".",
|
| 19 |
torch_dtype=torch.float16
|
| 20 |
).to("cuda")
|
| 21 |
-
|
| 22 |
pipe.enable_xformers_memory_efficient_attention()
|
| 23 |
print("Model loaded.")
|
| 24 |
|
| 25 |
|
| 26 |
-
# ===== Helpers =====
|
| 27 |
-
|
| 28 |
def decode_image(b64_string):
|
| 29 |
image_bytes = base64.b64decode(b64_string)
|
| 30 |
return Image.open(io.BytesIO(image_bytes)).convert("RGB")
|
|
@@ -42,32 +39,28 @@ def frames_to_base64_mp4(frames):
|
|
| 42 |
return base64.b64encode(video_bytes).decode()
|
| 43 |
|
| 44 |
|
| 45 |
-
# ===== ROUTES =====
|
| 46 |
-
|
| 47 |
@app.post("/video")
|
| 48 |
async def video_route(body: dict):
|
| 49 |
try:
|
| 50 |
image_b64 = body["image"]
|
| 51 |
prompt = body.get("prompt", "")
|
| 52 |
|
| 53 |
-
# Decode image
|
| 54 |
image = decode_image(image_b64)
|
| 55 |
|
| 56 |
-
# Run WAN model
|
| 57 |
output = pipe(image=image, prompt=prompt)
|
| 58 |
frames = output.frames
|
| 59 |
|
| 60 |
-
# Convert frames → mp4 → base64
|
| 61 |
video_b64 = frames_to_base64_mp4(frames)
|
| 62 |
|
| 63 |
return JSONResponse({"video": video_b64})
|
|
|
|
| 64 |
except Exception as e:
|
| 65 |
return JSONResponse({"error": str(e)}, status_code=500)
|
| 66 |
|
| 67 |
|
| 68 |
-
# HF Spaces launch
|
| 69 |
def start():
|
| 70 |
uvicorn.run(app, host="0.0.0.0", port=7860)
|
| 71 |
|
|
|
|
| 72 |
if __name__ == "__main__":
|
| 73 |
start()
|
|
|
|
| 10 |
import torch
|
| 11 |
from custom_wan_pipeline import WanImageToVideoPipeline
|
| 12 |
|
| 13 |
+
|
| 14 |
app = FastAPI()
|
| 15 |
|
|
|
|
| 16 |
print("Loading WAN I2V model...")
|
| 17 |
pipe = WanImageToVideoPipeline.from_pretrained(
|
| 18 |
+
".",
|
| 19 |
torch_dtype=torch.float16
|
| 20 |
).to("cuda")
|
|
|
|
| 21 |
pipe.enable_xformers_memory_efficient_attention()
|
| 22 |
print("Model loaded.")
|
| 23 |
|
| 24 |
|
|
|
|
|
|
|
| 25 |
def decode_image(b64_string):
|
| 26 |
image_bytes = base64.b64decode(b64_string)
|
| 27 |
return Image.open(io.BytesIO(image_bytes)).convert("RGB")
|
|
|
|
| 39 |
return base64.b64encode(video_bytes).decode()
|
| 40 |
|
| 41 |
|
|
|
|
|
|
|
| 42 |
@app.post("/video")
|
| 43 |
async def video_route(body: dict):
|
| 44 |
try:
|
| 45 |
image_b64 = body["image"]
|
| 46 |
prompt = body.get("prompt", "")
|
| 47 |
|
|
|
|
| 48 |
image = decode_image(image_b64)
|
| 49 |
|
|
|
|
| 50 |
output = pipe(image=image, prompt=prompt)
|
| 51 |
frames = output.frames
|
| 52 |
|
|
|
|
| 53 |
video_b64 = frames_to_base64_mp4(frames)
|
| 54 |
|
| 55 |
return JSONResponse({"video": video_b64})
|
| 56 |
+
|
| 57 |
except Exception as e:
|
| 58 |
return JSONResponse({"error": str(e)}, status_code=500)
|
| 59 |
|
| 60 |
|
|
|
|
| 61 |
def start():
|
| 62 |
uvicorn.run(app, host="0.0.0.0", port=7860)
|
| 63 |
|
| 64 |
+
|
| 65 |
if __name__ == "__main__":
|
| 66 |
start()
|