Paste a YouTube URL or upload an audio file. We'll separate instruments, remove noise, correct EQ for the recording era, and master the output for modern sound systems.
Listen to examples in the Showcase tab →
Sign in with Google to track your jobs across devices
49 restored recordings from the golden age of tango
We restore 1920s–1940s Argentine tango recordings for use by tango DJs (musicalizadores) at milongas. The goal is to take noisy, degraded vintage recordings and make them sound clear, warm, and ready for modern PA systems — while preserving the musical character of the original era.
Each recording goes through a fully automated ML restoration process:
Total processing time is approximately 12 times the track length (e.g., ~40 minutes for a 3-minute track). The ML denoising is fully automatic with no parameters to tune. Postprocessing adjustments take only seconds.
Every restored track is evaluated across 5 dimensions of audio quality, weighted to reflect what matters most for tango DJs: noise removal accounts for 65% of the score, because a clean recording is the single biggest improvement for playback at a milonga.
| Dimension | Weight | What It Measures |
|---|---|---|
| Clarity | 40% | How clean the audio is — absence of hiss, crackle, and background noise. Bell-curved: penalizes both noise AND over-cleaning. |
| Transients | 25% | Instrument attack clarity — whether you can hear the bandoneon staccato, violin phrasing, and piano accents that dancers respond to. |
| Warmth | 20% | Low-mid frequency body — the tonal richness from strings, bandoneon, and bass that gives tango its emotional weight. |
| Naturalness | 10% | Preservation of natural instrument texture — detects when denoising strips the aperiodic energy (breath, bow noise) that gives instruments their life. |
| Consistency | 5% | Volume stability over time — steady loudness makes it easier for DJs to mix tracks in a tanda without riding the fader. |
The ML denoising is fully automatic — no parameters to tune. After processing, you can adjust these optional postprocessing settings instantly without waiting:
| Option | What It Does | Default |
|---|---|---|
| Stereo Width | Adds spatial width to mono recordings using the Haas effect. Mono-compatible for dance floor PA systems. | 100% (standard width) |
| Loudness | Target output level in LUFS (broadcast standard). Lower = quieter, higher = louder. | −16 LUFS |
| Compression | Reduces volume range for consistent playback. Off by default — add if you need even volume for mixing in tandas. | Off (1.0:1) |
| Exciter | Adds high-frequency brightness via harmonic generation. Can bring sparkle to dull recordings. | Off |
| Harmonics | Pitch-tracked restoration of missing overtones. Synthesizes back instrument harmonics that were lost. Try Medium or High for very noisy sources. | Off |
For questions, feedback, or feature requests, please reach out via email or GitHub.
If you believe content processed through this service infringes your copyright, please contact us with details of the original work and the specific content in question. We will respond promptly.
Tango Restoration is a service for restoring historical tango recordings using AI-powered noise removal and audio refinement. Built for tango DJs (musicalizadores) who need clean, high-quality audio for milongas.
About 12 times the track length (e.g., ~40 minutes for a 3-minute track). The AI noise removal and audio refinement run on CPU. Postprocessing adjustments (stereo, loudness, compression) are instant — no waiting. Jobs are queued and processed one at a time.
Yes! Once a track is processed, the AI result is cached. Postprocessing options (stereo, loudness, compression, harmonics) can be changed instantly using the “Tweak Settings” button — no waiting at all.
Some noise on 78rpm shellac records is physically embedded in the music signal itself. Our AI removes the vast majority of surface noise, hiss, and crackle, but a small amount of residual noise may remain on very old or heavily worn recordings. This is a physical limitation of the original medium.
You can paste a YouTube URL or upload MP3, WAV, FLAC, M4A, OGG, or OPUS files (max 50 MB, 15 minutes). Output is 320kbps MP3.
Processed files are automatically deleted after 3 days. Download your restored tracks promptly.
The AI noise removal works on any historical recording (jazz, blues, classical). The model was trained specifically on tango, so it performs best on that genre, but it often produces good results on other music from the same era.
Each track is evaluated across 5 dimensions: Clarity, Transients, Warmth, Naturalness, and Consistency. The scores help compare before and after, but they are not perfect — trust your ears over the numbers.