VibeVoice

Public scan — anyone with this URL can view this analysis. Sign up to track your own repos privately, run scheduled re-scans, and get AI fix prompts via your dashboard.

62 of your 86 findings came from Repobility's proprietary detections. ✓ Repobility tags below mark them.

Upstream (GitHub) caused delay on this scan — not Repobility.

GitHub API rate-limited (HTTP 403) — preflight skipped, fell back to direct git clone.
Clone from GitHub took 132.35s for a 149.4 MB repo slow.
Repobility's analysis ran in 1.12s after the clone landed.

https://github.com/microsoft/VibeVoice.git · scanned 2026-05-24 01:24 UTC (1 week, 5 days ago) · 10 languages

216 findings (80 legacy + 136 scanner) 11/13 scanners ran 35th percentile · Python · small (2-20K LoC) Scanner says 92 (lower by 28)

File as issue Image v2 schema

UNIFIED Repobility · multi-layer engine · AI coders

Complete repo analysis

Last scanned 1 week, 5 days ago · v2 · 148 findings from 2 sources. Findings combine the legacy security pipeline AND the multi-layer engine (atlas, wiring, flows, ranked) AND verified AI agent contributions.

JSON

Score breakdown â 2026-05-18-v5

Component	Sub-score	Weight	Contribution
`structure_score`	75.0	0.15	11.25
`security_score`	100.0	0.25	25.00
`testing_score`	13.0	0.20	2.60
`documentation_score`	95.0	0.15	14.25
`practices_score`	40.0	0.15	6.00
`code_quality`	45.0	0.10	4.50
Overall		1.00	63.6

security_score may be inflated — optional security scanners were skipped on this fast scan

Severity distribution — click a segment to filter

Active filters: excluding tests × Reset all

Severity: Critical 0 High 69 Medium 16 Low 54 9-Layer: Software Security Quality Integrity Frontend Hardware Data Network Cicd Source: Legacy 80 9-layer 68 Crowd 0 Layer: Quality 108 Security 13 Software 23 Frontend 1 Cicd 1 Api 2

Exclude dismissed / FP Exclude test files

All 547 nodes from the latest scan, grouped by kind. Each node is a unit the engine identified (file, function, endpoint, table…). Most users won't need this view — it's primarily for debugging the engine's graph extraction or for AI agents that want to enumerate the project structure.

Label	Layer	Status	Path
`register_vibevoice`	software	dead	`vllm_plugin/__init__.py:20`
`_ffmpeg_load_bytes`	software	healthy	`vllm_plugin/model.py:26`
`_ffmpeg_load_file`	software	healthy	`vllm_plugin/model.py:37`
`load_bytes`	software	dead	`vllm_plugin/model.py:60`
`load_base64`	software	dead	`vllm_plugin/model.py:63`
`load_file`	software	dead	`vllm_plugin/model.py:66`
`__init__`	software	healthy	`vllm_plugin/model.py:950`
`forward`	software	dead	`vllm_plugin/model.py:1202`
`get_cfg`	software	healthy	`vllm_plugin/model.py:186`
`_ensure_audio_encoder_dtype`	software	healthy	`vllm_plugin/model.py:272`
`_iter_segments`	software	healthy	`vllm_plugin/model.py:391`
`get_hf_config`	software	healthy	`vllm_plugin/model.py:480`
`get_feature_extractor`	software	dead	`vllm_plugin/model.py:483`
`get_audio_token_info`	software	healthy	`vllm_plugin/model.py:531`
`get_supported_mm_limits`	software	dead	`vllm_plugin/model.py:556`
`get_mm_max_tokens_per_item`	software	dead	`vllm_plugin/model.py:559`
`_cfg`	software	healthy	`vllm_plugin/model.py:616`
`_get_max_audio_samples`	software	healthy	`vllm_plugin/model.py:608`
`get_dummy_text`	software	healthy	`vllm_plugin/model.py:632`
`get_dummy_mm_data`	software	healthy	`vllm_plugin/model.py:641`
`get_dummy_processor_inputs`	software	dead	`vllm_plugin/model.py:666`
`_vibevoice_field_config`	software	healthy	`vllm_plugin/model.py:679`
`_get_data_parser`	software	healthy	`vllm_plugin/model.py:710`
`_call_hf_processor`	software	healthy	`vllm_plugin/model.py:716`
`_hf_processor_applies_updates`	software	healthy	`vllm_plugin/model.py:787`
`_get_mm_fields_config`	software	healthy	`vllm_plugin/model.py:800`
`_get_prompt_updates`	software	healthy	`vllm_plugin/model.py:808`
`_tok_id`	software	healthy	`vllm_plugin/model.py:831`
`_to_int_len`	software	healthy	`vllm_plugin/model.py:866`
`get_replacement`	software	dead	`vllm_plugin/model.py:877`
`get_placeholder_str`	software	healthy	`vllm_plugin/model.py:938`
`compute_logits`	software	healthy	`vllm_plugin/model.py:984`
`embed_multimodal`	software	dead	`vllm_plugin/model.py:988`
`flatten_lengths`	software	healthy	`vllm_plugin/model.py:1010`
`get_input_embeddings`	software	healthy	`vllm_plugin/model.py:1103`
`embed_input_ids`	software	dead	`vllm_plugin/model.py:1127`
`get_language_model`	software	dead	`vllm_plugin/model.py:1166`
`load_weights`	software	healthy	`vllm_plugin/model.py:1170`
`load_audio`	software	healthy	`vllm_plugin/inputs.py:23`
`vibevoice_audio_input_mapper`	software	dead	`vllm_plugin/inputs.py:43`
`_guess_mime_type`	software	healthy	`vllm_plugin/tests/test_api_auto_recover.py:47`
`_get_duration_seconds_ffprobe`	software	healthy	`vllm_plugin/tests/test_api_auto_recover.py:64`
`_extract_audio_from_video`	software	healthy	`vllm_plugin/tests/test_api_auto_recover.py:75`
`_is_video_file`	software	healthy	`vllm_plugin/tests/test_api_auto_recover.py:96`
`_find_last_segment_boundary`	software	healthy	`vllm_plugin/tests/test_api_auto_recover.py:102`
`_find_safe_print_boundary`	software	healthy	`vllm_plugin/tests/test_api_auto_recover.py:114`
`__init__`	software	healthy	`vllm_plugin/tests/test_api_auto_recover.py:129`
`add_text`	software	healthy	`vllm_plugin/tests/test_api_auto_recover.py:138`
`_check_repetition`	software	healthy	`vllm_plugin/tests/test_api_auto_recover.py:143`
`_is_meaningful`	software	healthy	`vllm_plugin/tests/test_api_auto_recover.py:201`

Showing first 50 of this kind. Full payload available via the JSON button at the top of the page.

Label	Layer	Status	Path
`_PatchedAudioMediaIO`	software	healthy	`vllm_plugin/model.py:57`
`SpeechConnector`	software	healthy	`vllm_plugin/model.py:135`
`LlamaRMSNorm`	software	healthy	`vllm_plugin/model.py:153`
`VibeVoiceAudioEncoder`	software	healthy	`vllm_plugin/model.py:168`
`VibeVoiceProcessingInfo`	software	healthy	`vllm_plugin/model.py:477`
`VibeVoiceDummyInputsBuilder`	software	healthy	`vllm_plugin/model.py:596`
`VibeVoiceMultiModalProcessor`	software	healthy	`vllm_plugin/model.py:702`
`VibeVoiceForCausalLM`	software	healthy	`vllm_plugin/model.py:929`
`RepetitionDetector`	software	healthy	`vllm_plugin/tests/test_api_auto_recover.py:126`
`VibeVoiceAPIClient`	software	healthy	`vllm_plugin/scripts/gradio_asr_demo_api_video.py:698`
`VibeVoiceASRInference`	software	healthy	`demo/vibevoice_asr_gradio_demo.py:59`
`StopOnFlag`	software	healthy	`demo/vibevoice_asr_gradio_demo.py:439`
`VoiceMapper`	software	healthy	`demo/realtime_model_inference_from_file.py:21`
`VibeVoiceASRBatchInference`	software	healthy	`demo/vibevoice_asr_inference_from_file.py:26`
`StreamingTTSService`	software	healthy	`demo/web/app.py:43`
`UniformSampler`	software	healthy	`vibevoice/schedule/timestep_sampler.py:5`
`LogitNormalSampler`	software	healthy	`vibevoice/schedule/timestep_sampler.py:11`
`DPMSolverMultistepScheduler`	software	healthy	`vibevoice/schedule/dpm_solver.py:122`
`VibeVoiceTokenizerProcessor`	software	healthy	`vibevoice/processor/vibevoice_tokenizer_processor.py:21`
`VibeVoiceASRProcessor`	software	healthy	`vibevoice/processor/vibevoice_asr_processor.py:30`
`AudioNormalizer`	software	healthy	`vibevoice/processor/audio_utils.py:149`
`VibeVoiceStreamingProcessor`	software	healthy	`vibevoice/processor/vibevoice_streaming_processor.py:17`
`VibeVoiceProcessor`	software	healthy	`vibevoice/processor/vibevoice_processor.py:17`
`VibeVoiceStreamingConfig`	software	healthy	`vibevoice/modular/configuration_vibevoice_streaming.py:14`
`VibeVoiceASRPreTrainedModel`	software	healthy	`vibevoice/modular/modeling_vibevoice_asr.py:30`
`VibeVoiceASRModel`	software	healthy	`vibevoice/modular/modeling_vibevoice_asr.py:62`
`VibeVoiceASRForConditionalGeneration`	software	healthy	`vibevoice/modular/modeling_vibevoice_asr.py:152`
`RMSNorm`	software	healthy	`vibevoice/modular/modular_vibevoice_diffusion_head.py:20`
`TimestepEmbedder`	software	healthy	`vibevoice/modular/modular_vibevoice_diffusion_head.py:48`
`FeedForwardNetwork`	software	healthy	`vibevoice/modular/modular_vibevoice_diffusion_head.py:96`
`HeadLayer`	software	healthy	`vibevoice/modular/modular_vibevoice_diffusion_head.py:126`
`FinalLayer`	software	healthy	`vibevoice/modular/modular_vibevoice_diffusion_head.py:164`
`VibeVoiceDiffusionHead`	software	healthy	`vibevoice/modular/modular_vibevoice_diffusion_head.py:191`
`BinaryClassifier`	software	healthy	`vibevoice/modular/modeling_vibevoice_streaming.py:32`
`SpeechConnector`	software	healthy	`vibevoice/modular/modeling_vibevoice_streaming.py:44`
`VibeVoiceStreamingPreTrainedModel`	software	healthy	`vibevoice/modular/modeling_vibevoice_streaming.py:59`
`VibeVoiceStreamingModel`	software	healthy	`vibevoice/modular/modeling_vibevoice_streaming.py:94`
`VibeVoiceTextTokenizer`	software	healthy	`vibevoice/modular/modular_vibevoice_text_tokenizer.py:12`
`VibeVoiceTextTokenizerFast`	software	healthy	`vibevoice/modular/modular_vibevoice_text_tokenizer.py:112`
`VibeVoiceASRTextTokenizerFast`	software	healthy	`vibevoice/modular/modular_vibevoice_text_tokenizer.py:210`
`VibeVoiceAcousticTokenizerConfig`	software	healthy	`vibevoice/modular/configuration_vibevoice.py:31`
`VibeVoiceSemanticTokenizerConfig`	software	healthy	`vibevoice/modular/configuration_vibevoice.py:94`
`VibeVoiceDiffusionHeadConfig`	software	healthy	`vibevoice/modular/configuration_vibevoice.py:148`
`VibeVoiceConfig`	software	healthy	`vibevoice/modular/configuration_vibevoice.py:182`
`VibeVoiceASRConfig`	software	healthy	`vibevoice/modular/configuration_vibevoice.py:289`
`MockCacheLayer`	software	healthy	`vibevoice/modular/modeling_vibevoice_streaming_inference.py…`
`VibeVoiceCausalLMOutputWithPast`	software	healthy	`vibevoice/modular/modeling_vibevoice_streaming_inference.py…`
`VibeVoiceGenerationOutput`	software	healthy	`vibevoice/modular/modeling_vibevoice_streaming_inference.py…`
`VibeVoiceStreamingForConditionalGenerationInference`	software	healthy	`vibevoice/modular/modeling_vibevoice_streaming_inference.py…`
`VibeVoiceCausalLMOutputWithPast`	software	healthy	`vibevoice/modular/modeling_vibevoice.py:34`

Showing first 50 of this kind. Full payload available via the JSON button at the top of the page.

Label	Layer	Status	Path
`README.md`	software	healthy	`README.md`
`CONTRIBUTING.md`	software	healthy	`CONTRIBUTING.md`
`pyproject.toml`	software	healthy	`pyproject.toml`
`SECURITY.md`	software	healthy	`SECURITY.md`
`__init__.py`	software	healthy	`vllm_plugin/__init__.py`
`model.py`	software	healthy	`vllm_plugin/model.py`
`inputs.py`	software	healthy	`vllm_plugin/inputs.py`
`test_api_auto_recover.py`	software	healthy	`vllm_plugin/tests/test_api_auto_recover.py`
`test_api.py`	software	healthy	`vllm_plugin/tests/test_api.py`
`start_server.py`	software	healthy	`vllm_plugin/scripts/start_server.py`
`gradio_asr_demo_api_video.py`	software	healthy	`vllm_plugin/scripts/gradio_asr_demo_api_video.py`
`generate_tokenizer_files.py`	software	healthy	`vllm_plugin/tools/generate_tokenizer_files.py`
`download_experimental_voices.sh`	software	healthy	`demo/download_experimental_voices.sh`
`vibevoice_asr_gradio_demo.py`	software	healthy	`demo/vibevoice_asr_gradio_demo.py`
`realtime_model_inference_from_file.py`	software	healthy	`demo/realtime_model_inference_from_file.py`
`vibevoice_asr_inference_from_file.py`	software	healthy	`demo/vibevoice_asr_inference_from_file.py`
`vibevoice_realtime_demo.py`	software	healthy	`demo/vibevoice_realtime_demo.py`
`app.py`	software	healthy	`demo/web/app.py`
`index.html`	software	healthy	`demo/web/index.html`
`vibevoice-tts.md`	software	healthy	`docs/vibevoice-tts.md`
`vibevoice-asr.md`	software	healthy	`docs/vibevoice-asr.md`
`vibevoice-realtime-0.5b.md`	software	healthy	`docs/vibevoice-realtime-0.5b.md`
`setup_gradio_demo.md`	software	healthy	`docs/setup_gradio_demo.md`
`vibevoice-vllm-asr.md`	software	healthy	`docs/vibevoice-vllm-asr.md`
`__init__.py`	software	healthy	`vibevoice/__init__.py`
`__init__.py`	software	healthy	`vibevoice/schedule/__init__.py`
`timestep_sampler.py`	software	healthy	`vibevoice/schedule/timestep_sampler.py`
`dpm_solver.py`	software	healthy	`vibevoice/schedule/dpm_solver.py`
`__init__.py`	software	healthy	`vibevoice/processor/__init__.py`
`vibevoice_tokenizer_processor.py`	software	healthy	`vibevoice/processor/vibevoice_tokenizer_processor.py`
`vibevoice_asr_processor.py`	software	healthy	`vibevoice/processor/vibevoice_asr_processor.py`
`audio_utils.py`	software	healthy	`vibevoice/processor/audio_utils.py`
`vibevoice_streaming_processor.py`	software	healthy	`vibevoice/processor/vibevoice_streaming_processor.py`
`vibevoice_processor.py`	software	healthy	`vibevoice/processor/vibevoice_processor.py`
`__init__.py`	software	healthy	`vibevoice/scripts/__init__.py`
`configuration_vibevoice_streaming.py`	software	healthy	`vibevoice/modular/configuration_vibevoice_streaming.py`
`__init__.py`	software	healthy	`vibevoice/modular/__init__.py`
`modeling_vibevoice_asr.py`	software	healthy	`vibevoice/modular/modeling_vibevoice_asr.py`
`modular_vibevoice_diffusion_head.py`	software	healthy	`vibevoice/modular/modular_vibevoice_diffusion_head.py`
`modeling_vibevoice_streaming.py`	software	healthy	`vibevoice/modular/modeling_vibevoice_streaming.py`
`modular_vibevoice_text_tokenizer.py`	software	healthy	`vibevoice/modular/modular_vibevoice_text_tokenizer.py`
`configuration_vibevoice.py`	software	healthy	`vibevoice/modular/configuration_vibevoice.py`
`modeling_vibevoice_streaming_inference.py`	software	healthy	`vibevoice/modular/modeling_vibevoice_streaming_inference.py`
`modeling_vibevoice.py`	software	healthy	`vibevoice/modular/modeling_vibevoice.py`
`streamer.py`	software	healthy	`vibevoice/modular/streamer.py`
`modular_vibevoice_tokenizer.py`	software	healthy	`vibevoice/modular/modular_vibevoice_tokenizer.py`
`qwen2.5_7b_32k.json`	software	healthy	`vibevoice/configs/qwen2.5_7b_32k.json`
`qwen2.5_1.5b_64k.json`	software	healthy	`vibevoice/configs/qwen2.5_1.5b_64k.json`
`inference_lora.py`	software	healthy	`finetuning-asr/inference_lora.py`
`README.md`	software	healthy	`finetuning-asr/README.md`

Showing first 50 of this kind. Full payload available via the JSON button at the top of the page.

Label	Layer	Status	Path
`vllm_plugin`	software	healthy	`vllm_plugin`
`tests`	software	healthy	`vllm_plugin/tests`
`scripts`	software	healthy	`vllm_plugin/scripts`
`tools`	software	healthy	`vllm_plugin/tools`
`demo`	software	healthy	`demo`
`web`	software	healthy	`demo/web`
`docs`	software	healthy	`docs`
`vibevoice`	software	healthy	`vibevoice`
`schedule`	software	healthy	`vibevoice/schedule`
`processor`	software	healthy	`vibevoice/processor`
`scripts`	software	healthy	`vibevoice/scripts`
`modular`	software	healthy	`vibevoice/modular`
`configs`	software	healthy	`vibevoice/configs`
`finetuning-asr`	software	healthy	`finetuning-asr`
`toy_dataset`	software	healthy	`finetuning-asr/toy_dataset`

Label	Layer	Status	Path
`GET /`	api	healthy	`demo/web/app.py`
`GET /config`	api	healthy	`demo/web/app.py`

Label	Layer	Status	Path
`nginx`	network	healthy	`vllm_plugin/scripts/start_server.py`
`loadbalancer`	network	healthy	`vllm_plugin/scripts/start_server.py`

Label	Layer	Status	Path
`repobility-clone-0t7lvqoy`	software	healthy	`/tmp/repobility-clone-0t7lvqoy`

Label	Layer	Status	Path
`127.0.0.1`	network	healthy	`vllm_plugin/scripts/start_server.py`

Label	Layer	Status	Path
`gpu (detected)`	hardware	healthy	`vllm_plugin/model.py`

Label	Layer	Status	Path
`vps::azure`	hardware	healthy	`vllm_plugin/tests/test_api_auto_recover.py`

For AI agents: Voting guide (TP/FP) MCP manifest Stdio wrapper SARIF Integrate Findings queue Vote TP/FP on findings to calibrate the engine.

For AI agents + API integrations

Email me when this repo regresses

Free. We re-scan periodically; new criticals → your inbox. No signup required for the scan itself.

API access

This page is publicly accessible at: https://repobility.com/scan/59822133-99df-4890-8acc-0ec11ec6560c/

To check status programmatically (no auth required):

curl -s https://repobility.com/api/v1/public/scan/59822133-99df-4890-8acc-0ec11ec6560c/

Important — please don't re-submit the same URL repeatedly. The submission endpoint is idempotent: re-submitting the same git URL returns this same scan_token, not a new one. To re-scan this repo, sign up free and use the dashboard.

VibeVoice

Complete repo analysis

function 50+

class 50+

file 50+

directory 15

endpoint 2

load_balancer 2

repo 1

service 1

gpu 1

vps 1