[BugFix] Standardize StableAudio audio output (#842)

Signed-off-by: LudovicoYIN <hankeyin@gmail.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
This commit is contained in:
YinHankeandHongsheng Liu authored and GitHub committed 2026-01-24 22:27:58 +13:00
1 parent 3d9cd16df7
commit f9c69a8c40
6 files changed
+90 -30

No files matched your search

@@ -142,7 +142,7 @@ def main():
generation_start = time.perf_counter()
# Generate audio
audio = omni.generate(
outputs = omni.generate(
args.prompt,
negative_prompt=args.negative_prompt,
generator=generator,
@@ -166,6 +166,21 @@ def main():
suffix = output_path.suffix or ".wav"
stem = output_path.stem or "stable_audio_output"
# Extract audio from omni.generate() outputs
if not outputs:
raise ValueError("No output generated from omni.generate()")
output = outputs[0]
if not hasattr(output, "request_output") or not output.request_output:
raise ValueError("No request_output found in OmniRequestOutput")
request_output = output.request_output[0]
if not hasattr(request_output, "multimodal_output"):
raise ValueError("No multimodal_output found in request_output")
audio = request_output.multimodal_output.get("audio")
if audio is None:
raise ValueError("No audio output found in request_output")
# Handle different output formats
if isinstance(audio, torch.Tensor):
audio = audio.cpu().float().numpy()