Funasr using hugging face hub with paraformer-zh errors

A bug in FunASR's paraformer-zh model on Hugging Face Hub causes missing timestamps and a KeyError: 'text' error, according to a developer's investigation. The issue can be worked around by setting pred_timestamp=True, which restored timestamps in controlled tests on CPU and T4. The problem appears to be two separate timestamp-related bugs in FunASR rather than a fundamental incompatibility with the intended pipeline.

For now, after looking into this, it seems quite likely that the issue is related to a bug on the FunASR side : I do not think your intended setup is fundamentally unsupported. The current funasr/paraformer-zh Hugging Face model card https://huggingface.co/funasr/paraformer-zh explicitly shows the full pipeline with VAD, punctuation, and speaker diarization, and says that the output includes timestamps and speaker labels. So I would not abandon the HF route or the diarization goal just because of this error. What I found is closer to two separate timestamp-related problems in FunASR , plus a secondary output-schema issue. The cheapest first test is simply: result = model.generate input=audio, pred timestamp=True, In controlled reproductions with the HF Paraformer model, this was enough to change the ASR result from no timestamp to timestamp present , on both CPU and T4. So, for practical debugging, I would start there before changing the model stack. A rough decision flow is: add pred timestamp=True | +-- still no timestamp | - check which model/config actually resolved | +-- timestamp appears | +-- timestamps stay within the audio duration | - continue checking the speaker pipeline | +-- timestamps extend well beyond the audio duration - likely the separate GPU dynamic-batching issue below There is also a public reproduction notebook here: 1. Why the missing timestamp looks like an internal FunASR integration issueSo my default route would be: pred timestamp=True KeyError: 'text' after missing timestamps as likely secondary, and inspect the actual sentence info keys rather than assuming the documented schema was reached.So, at least from these reproductions, this looks much more like a couple of FunASR integration/length-handling bugs that can be separated and worked around than a reason to give up on what you are trying to build.