81
Stack Score
Best for
Developers building speech-to-text pipelines who want the accuracy benchmark and the option to self-host.
At a glance
- Pricing
- free
- Setup time
- 1 hour
- Learning curve
- medium
- Last verified
- 5/19/2026
What we love
- ✓Free forever if you self-host — no API costs
- ✓Accuracy is the industry baseline that competitors are measured against
- ✓Massive ecosystem of wrappers, finetunes, and integrations
Where it falls short
- −No streaming — input must be a full audio file (work around with VAD)
- −No speaker diarization out of the box (need pyannote or similar)
- −Hallucinates on long silences or background music
Pricing
✓ Pricing verified Jun 21, 20262 tiers
Self-hosted
$0
MIT-style open weights
- ✓Run on your hardware
- ✓Full offline use
- ✓No usage limits
OpenAI API
$0.006
per minute of audio
- ✓Hosted endpoint
- ✓Large-v2 model
- ✓JSON or SRT output
Why we picked it
The de facto standard for any speech-to-text pipeline.
Overview
OpenAI's open-source ASR model. Free to run locally, paid via OpenAI API. Multilingual, robust to accents and noise.
Key features
- ●Open-weights model — run locally on consumer hardware
- ●99 languages supported with strong accent robustness
- ●Word-level timestamps for subtitling and editing workflows
- ●Available via OpenAI API at $0.006 per minute
- ●Large-v3 model is the current accuracy benchmark for ASR
- ●Faster-whisper and whisper.cpp ports run on CPU and Apple Silicon
Best use cases
Podcast and video transcription
Generate searchable transcripts and subtitle files from any recorded audio.
Meeting note pipelines
Pair with an LLM to turn a 60-minute recording into action items and summaries.
Multilingual content workflows
Transcribe and translate non-English audio in one pass for global teams.
Integrations
Similar tools
See all voice & audio tools →One AI Stack pick every week.
The tool we’d pay for this week, why we’d pay for it, and what we’d skip. No promo spam.