Update README.md
Browse files
README.md
CHANGED
|
@@ -655,4 +655,37 @@ Its goal is to let users explore sound in new ways while retaining precise contr
|
|
| 655 |
- how it feels sonically
|
| 656 |
- how it fits into a production workflow
|
| 657 |
|
| 658 |
-
That combination of **musical structure**, **instrument identity**, **timbral control**, and **loop fidelity** is what defines the model.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 655 |
- how it feels sonically
|
| 656 |
- how it fits into a production workflow
|
| 657 |
|
| 658 |
+
That combination of **musical structure**, **instrument identity**, **timbral control**, and **loop fidelity** is what defines the model.
|
| 659 |
+
|
| 660 |
+
Code for running the weight in Diffusers
|
| 661 |
+
```
|
| 662 |
+
import scipy
|
| 663 |
+
import torch
|
| 664 |
+
import soundfile as sf
|
| 665 |
+
from diffusers import StableAudioPipeline
|
| 666 |
+
|
| 667 |
+
repo_id = "tintwotin/Foundation-1-Diffusers"
|
| 668 |
+
pipe = StableAudioPipeline.from_pretrained(repo_id, torch_dtype=torch.float16)
|
| 669 |
+
pipe = pipe.to("cuda")
|
| 670 |
+
|
| 671 |
+
# define the prompts
|
| 672 |
+
prompt = "Bass, FM Bass, Medium Delay, Medium Reverb, Low Distortion, Phaser, Sub Bass, Bass, Upper Mids, Acid, Gritty, Wide, Dubstep, Thick, Silky, Warm, Rich, Overdriven, Crisp, Deep, Clean, Pitch Bend, 303, 8 Bars, 140 BPM, E minor"
|
| 673 |
+
negative_prompt = "Low quality."
|
| 674 |
+
|
| 675 |
+
# set the seed for generator
|
| 676 |
+
generator = torch.Generator("cuda").manual_seed(0)
|
| 677 |
+
|
| 678 |
+
# run the generation
|
| 679 |
+
audio = pipe(
|
| 680 |
+
prompt,
|
| 681 |
+
negative_prompt=negative_prompt,
|
| 682 |
+
num_inference_steps=200,
|
| 683 |
+
audio_end_in_s=10.0,
|
| 684 |
+
num_waveforms_per_prompt=1,
|
| 685 |
+
generator=generator,
|
| 686 |
+
).audios
|
| 687 |
+
|
| 688 |
+
output = audio[0].T.float().cpu().numpy()
|
| 689 |
+
sf.write("./foundation_loop.wav", output, pipe.vae.sampling_rate)
|
| 690 |
+
|
| 691 |
+
```
|