Whisperx
App in the BluixApps catalog
What it is
WhisperX is the production-grade enhancement of OpenAI Whisper — adds 70× real-time inference speed, word-level timestamps via forced alignment, and speaker diarization via pyannote-audio. The standard for serious transcription + subtitling pipelines.
Where vanilla Whisper is research code, WhisperX is the engineering-grade version.
What it's for
- Audio/video transcription at production speed
- Multi-speaker labelling (who said what)
- Word-level timestamps for precise subtitle generation
- VAD pre-processing — skip silence, faster results
- Batch processing — entire podcasts, full meetings
- Multilingual — 99 languages from Whisper backbone
Who it's for
- Podcast producers generating multi-speaker transcripts
- Video platforms auto-generating SRT/VTT subtitles
- Call center analytics transcribing customer calls at scale
- Meeting note systems (Zoom transcripts, Teams summaries)
- Accessibility teams captioning content
- AI app developers building voice-to-text pipelines
Why teams pick WhisperX over alternatives
- BSD-4-Clause — fully open
- Faster than vanilla Whisper by 10-70× (batched + VAD + faster-whisper backend)
- Word-level timestamps — forced alignment via wav2vec2
- Diarization via pyannote-audio integration
- Production-tested by major podcasts, transcription services
- Better than commercial APIs for many use cases at zero per-minute cost
Integrations
- REST API server — POST audio → JSON/SRT/VTT
- Whisper backbone — uses faster-whisper or original Whisper
- pyannote-audio for speaker diarization
- VAD — Silero or pyannote VAD preprocessor
- Pair with: Ollama/vLLM for "transcribe → summarize" pipelines
- Output formats: JSON, SRT, VTT, TSV, TXT
Notable users & community
- 17k+ GitHub stars (parent WhisperX)
- Used by podcast platforms (Podscribe-style services)
- Featured in major transcription tooling roundups
- Active research community + commercial integrations
- Multiple production-tested deployments
Tips & operations
- VRAM: 6 GB GPU for large-v3 model; 4 GB CPU fallback works
- Speed: 70× real-time on RTX 4090; 10-30× on RTX 3060
- Diarization: requires HF_TOKEN + accept pyannote terms
- VAD preprocessing: enables "skip silence" — massive speed-up on podcasts
- Language detection: automatic or specify language for accuracy
- Output formats:
- JSON: full timestamps + speaker labels
- SRT/VTT: ready for video players
- TSV: tabular for spreadsheets
- Best languages: EN, ES, FR, DE, IT, PT, RU, ZH, JA
- Production: batch via API, async queue, file upload limit ~100 MB
What we ship in BluixApps
- Docker (ghcr.io/jim60105/whisperx:latest) with NVIDIA runtime
- Persistent volumes: audio (input), output (transcripts), models cache
- Port 8002 mapped
- HF_TOKEN environment variable for diarization
- Install report at
/root/bluixapps/whisperx.txt - Sample curl commands for transcription + diarization
- Output format selection guide
- Use case examples (podcast, video subtitles, meetings)
- Pairing suggestions (LLM for summarization)
- GPU pre-flight check via
bluixapps_ensure_nvidia_runtime - Backup hook covers audio + output + models cache
Get this app — pick a BluixApps plan
Same catalog. Scaling tenant isolation, white-label and support tier.
| Tier | Tenants | Catalog | Support | White-label | Monthly | |
|---|---|---|---|---|---|---|
| Stacks | 1 | 19 curated stacks | Standard | — | $19/mo | DetailDeploy |
| Starter | 10 | Full catalog | Standard | +$15–25/mo | $49/mo | DetailDeploy |
| Pro | 25 | Full catalog | Priority bugfix | +$15–25/mo | $149/mo | DetailDeploy |
| Growth | 100 | Full catalog | Priority bugfix | +$15–25/mo | $349/mo | DetailDeploy |
| Scale | 500 | Full catalog | 7-day window | +$15–25/mo | $799/mo | DetailDeploy |
| Enterprise | Unlimited | Full catalog | Priority 7-day | Bundled | $1,499/mo | DetailDeploy |
