Clean audio + transcription in one pass
SoundHalo cleans messy real-world recordings, from cafe interviews to old voice memos, then transcribes them into accurate text. Denoise and normalize with our cleanup tools, then get a transcript with timestamps, plus speaker labels when separate voices are detected.
Free Tools
Each one runs a step of the SoundHalo pipeline in your browser. Free to use, no signup.
Accurate transcripts with timestamps from any audio or video file.
Turn MP3 recordings into accurate, timestamped text.
Ready-to-use subtitle files with timestamps for any video.
Even out quiet and loud passages to a consistent loudness.
Strip hum, hiss, and room noise from messy recordings.
Check your microphone in the browser before you record.
Every tool is also an API. Read the docs →
How it works
Drop in any audio or video file. WAV, MP3, M4A, FLAC, OGG, MP4, and more all work.
SoundHalo removes background noise, evens out loudness, and trims dead silence automatically.
Get an accurate transcript with speaker labels and timestamps, in dozens of languages.
Download your transcript as TXT or SRT. The cleaned audio is yours to keep too.
Use Cases
Clean up remote guest audio and get timestamped transcripts for show notes and chapters.
Salvage noisy interview recordings and get accurate, speaker-labeled quotes you can trust.
Turn lecture-hall recordings into searchable notes, even when the room was loud.
Transcribe field recordings and focus groups with speakers separated for easy coding.
Clean up quick voice memos and get readable text you can drop into any doc.
Yes. The tools work without an account or credit card. Free usage is limited to files up to 100 MB and 30 files per hour, which covers most everyday recordings. If you need longer files, batch uploads, or higher limits, a paid plan lifts them.
Your uploads and results are private and tied to a token only you hold, not a public account. Files are automatically deleted after 30 days. We do not sell your data or use your recordings to train models.
Most transcribers feed messy audio straight to a speech model, so background noise and uneven levels turn into garbled text. SoundHalo cleans the recording first (denoise, normalize loudness, trim dead silence), then transcribes the cleaned audio. Better input means a more accurate transcript.
Transcription works across dozens of languages and detects the spoken language automatically. Every transcript includes timestamps, and speaker labels are added when the provider detects separate speakers.
On the free tier, files can be up to 100 MB and you can process up to 30 files per hour. Paid plans raise both the file size and the rate limits.
Yes. Every tool is backed by the same HTTP API you can call from your own code. See the docs at /docs for endpoints, request formats, and examples.