[Usage]: Transcription "Maximum clip duration (30s) exceeded

[digicontacts]

Your current environment

The output of `python collect_env.py`

PyTorch version: 2.5.1+cu124
Is debug build: False
CUDA used to build PyTorch: 12.4
ROCM used to build PyTorch: N/A

OS: Rocky Linux 9.4 (Blue Onyx) (x86_64)
GCC version: (GCC) 11.5.0 20240719 (Red Hat 11.5.0-2)
Clang version: Could not collect
CMake version: version 3.26.5
Libc version: glibc-2.34

Python version: 3.12.9 | packaged by Anaconda, Inc. | (main, Feb 6 2025, 18:56:27) [GCC 11.2.0] (64-bit runtime)
Python platform: Linux-5.14.0-427.37.1.el9_4.x86_64-x86_64-with-glibc2.34
Is CUDA available: True
CUDA runtime version: Could not collect
CUDA_MODULE_LOADING set to: LAZY
GPU models and configuration:
GPU 0: NVIDIA GeForce RTX 3060
GPU 1: NVIDIA GeForce RTX 3060

Nvidia driver version: 560.35.03
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

CPU:
Architecture : x86_64
Mode(s) opératoire(s) des processeurs : 32-bit, 64-bit
Tailles des adresses: 48 bits physical, 48 bits virtual
Boutisme : Little Endian
Processeur(s) : 24
Liste de processeur(s) en ligne : 0-23
Identifiant constructeur : AuthenticAMD
Identifiant constructeur du BIOS : Advanced Micro Devices, Inc.
Nom de modèle : AMD Ryzen 9 5900X 12-Core Processor
Nom de modèle BIOS : AMD Ryzen 9 5900X 12-Core Processor
Famille de processeur : 25
Modèle : 33
Thread(s) par cœur : 2
Cœur(s) par socket : 12
Socket(s) : 1
Révision : 2
Accroissement de fréquence : activé
CPU(s) scaling MHz: 100%
Vitesse maximale du processeur en MHz : 3700,0000
Vitesse minimale du processeur en MHz : 2200,0000
BogoMIPS : 7386.08
Drapeaux : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc cpuid extd_apicid aperfmperf rapl pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand lahf_lm cmp_legacy extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt tce topoext perfctr_core perfctr_nb bpext perfctr_llc mwaitx cpb cat_l3 cdp_l3 hw_pstate ssbd mba ibrs ibpb stibp vmmcall fsgsbase bmi1 avx2 smep bmi2 erms invpcid cqm rdt_a rdseed adx smap clflushopt clwb sha_ni xsaveopt xsavec xgetbv1 xsaves cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local clzero irperf xsaveerptr rdpru wbnoinvd arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid decodeassists pausefilter pfthreshold avic v_vmsave_vmload vgif v_spec_ctrl umip pku ospke vaes vpclmulqdq rdpid overflow_recov succor smca fsrm debug_swap
Cache L1d : 384 KiB (12 instances)
Cache L1i : 384 KiB (12 instances)
Cache L2 : 6 MiB (12 instances)
Cache L3 : 64 MiB (2 instances)
Nœud(s) NUMA : 1
Nœud NUMA 0 de processeur(s) : 0-23
Vulnérabilité Gather data sampling : Not affected
Vulnérabilité Itlb multihit : Not affected
Vulnérabilité L1tf : Not affected
Vulnérabilité Mds : Not affected
Vulnérabilité Meltdown : Not affected
Vulnérabilité Mmio stale data : Not affected
Vulnérabilité Retbleed : Not affected
Vulnérabilité Spec rstack overflow : Mitigation; Safe RET
Vulnérabilité Spec store bypass : Mitigation; Speculative Store Bypass disabled via prctl
Vulnérabilité Spectre v1 : Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnérabilité Spectre v2 : Mitigation; Retpolines, IBPB conditional, IBRS_FW, STIBP always-on, RSB filling, PBRSB-eIBRS Not affected
Vulnérabilité Srbds : Not affected
Vulnérabilité Tsx async abort : Not affected

Versions of relevant libraries:
[pip3] numpy==1.26.4
[pip3] nvidia-cublas-cu12==12.4.5.8
[pip3] nvidia-cuda-cupti-cu12==12.4.127
[pip3] nvidia-cuda-nvrtc-cu12==12.4.127
[pip3] nvidia-cuda-runtime-cu12==12.4.127
[pip3] nvidia-cudnn-cu12==9.1.0.70
[pip3] nvidia-cufft-cu12==11.2.1.3
[pip3] nvidia-curand-cu12==10.3.5.147
[pip3] nvidia-cusolver-cu12==11.6.1.9
[pip3] nvidia-cusparse-cu12==12.3.1.170
[pip3] nvidia-nccl-cu12==2.21.5
[pip3] nvidia-nvjitlink-cu12==12.4.127
[pip3] nvidia-nvtx-cu12==12.4.127
[pip3] pyzmq==26.3.0
[pip3] torch==2.5.1
[pip3] torchaudio==2.5.1
[pip3] torchvision==0.20.1
[pip3] transformers==4.49.0
[pip3] triton==3.1.0
[conda] numpy 1.26.4 pypi_0 pypi
[conda] nvidia-cublas-cu12 12.4.5.8 pypi_0 pypi
[conda] nvidia-cuda-cupti-cu12 12.4.127 pypi_0 pypi
[conda] nvidia-cuda-nvrtc-cu12 12.4.127 pypi_0 pypi
[conda] nvidia-cuda-runtime-cu12 12.4.127 pypi_0 pypi
[conda] nvidia-cudnn-cu12 9.1.0.70 pypi_0 pypi
[conda] nvidia-cufft-cu12 11.2.1.3 pypi_0 pypi
[conda] nvidia-curand-cu12 10.3.5.147 pypi_0 pypi
[conda] nvidia-cusolver-cu12 11.6.1.9 pypi_0 pypi
[conda] nvidia-cusparse-cu12 12.3.1.170 pypi_0 pypi
[conda] nvidia-nccl-cu12 2.21.5 pypi_0 pypi
[conda] nvidia-nvjitlink-cu12 12.4.127 pypi_0 pypi
[conda] nvidia-nvtx-cu12 12.4.127 pypi_0 pypi
[conda] pyzmq 26.3.0 pypi_0 pypi
[conda] torch 2.5.1 pypi_0 pypi
[conda] torchaudio 2.5.1 pypi_0 pypi
[conda] torchvision 0.20.1 pypi_0 pypi
[conda] transformers 4.49.0 pypi_0 pypi
[conda] triton 3.1.0 pypi_0 pypi
ROCM Version: Could not collect
Neuron SDK Version: N/A
vLLM Version: 0.7.3
vLLM Build Flags:
CUDA Archs: Not Set; ROCm: Disabled; Neuron: Disabled
GPU Topology:
GPU0 GPU1 CPU Affinity NUMA Affinity GPU NUMA ID
GPU0 X PHB 0-23 0 N/A
GPU1 PHB X 0-23 0 N/A

Legend:

X = Self
SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)
NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node
PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)
PXB = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)
PIX = Connection traversing at most a single PCIe bridge
NV# = Connection traversing a bonded set of # NVLinks

LD_LIBRARY_PATH=/usr/local/anaconda3/envs/vllm_V4/lib/python3.12/site-packages/cv2/../../lib64:
NCCL_CUMEM_ENABLE=0
TORCHINDUCTOR_COMPILE_THREADS=1
CUDA_MODULE_LOADING=LAZY

How would you like to use vllm

I use this code to transcript an audio file

`
import requests

with open("audio-samples/audio.mp3", "rb") as audio_file:
response = requests.post(
"http://localhost:8001/v1/audio/transcriptions",
files={"file": audio_file},
data={
"model": "openai/whisper-large-v3-turbo",
"language": "fr",
"response_format": "json",
"timestamp_granularities[]": ["word", "segment"]
}
)
print("Transcription :", response.text)
`

How can i by pass the limitation of 30s

because i got this reply:

Transcription : {"object":"error","message":"Maximum clip duration (30s) exceeded.","type":"BadRequestError","param":null,"code":40

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[DarkLight1337]

cc @NickLucche

[NickLucche]

How can i by pass the limitation of 30s

This is not currently possible. Since #13631 landed, I will look into implementing a naive chunking strategy for whisper too as soon as I have bandwidth.
Contributions are always welcome though!

[NickLucche]

Following up with the obvious, you can implement the chunking strategy in your own code and call vllm to do inference on multiple audio clips.
That way you can have a more refined solution (ie overlap the chunks, clip on silence..)

[digicontacts]

Thank you for the reply, I'll add this logic on my end with overlapping chunk to keep the context.

By the way, I have a follow-up question.

I'm using a Json output because it seems to me that the json_verbose output is not implemented, but even with the options
data={ “model": ‘openai/whisper-large-v3-turbo’, “language": ‘fr’, “prompt": ‘’, “response_format": ‘json’, “timestamp_granularities[]": [“word”, “segment”], “temperature": temperature }

I don't get the timestamps of the segments in the response.

Can you tell me what I'm doing wrong?

[NickLucche]

Nothing wrong, the Json response format is following this spec https://platform.openai.com/docs/api-reference/audio/json-object , so no timestamps. Also the notion of segments does imply chunking of some sort.

[digicontacts]

ah ok, did you know when verbose_json would be available ?

[MohamedAliRashad]

Any updates ?

[NickLucche]

I have just started looking into chunking. Will work on it whenever I have free bandwidth.

[Anderstask1]

@NickLucche Thanks for looking into this! A few quick questions about the planned chunking approach:

• Will it use fixed-size windows with overlap, silence detection or a different strategy?
• Any plans to support verbose_json timestamps natively?

[Anderstask1]

And do you plan to support a stream of audio?

[AlexBlack2202]

do you have any solotuion for this problem

[taimans-git]

A general chunking approach from vllm for audio files would also be great for us.