I've been loving the app. The audio restoration works amazing on my suno songs
SAM Audio Online: Split stems from full songs
Looking for a real SAM Audio workflow? Use the Meta stem splitter approach on full songs in stereo, not just short demo clips.
SELECT SEPARATION TYPE
Upload audio files
Drag & drop or click to browse (multiple files supported)
Supports: MP3, MPEG, MP2, MPGA, WAV, WAV, AIFF, AIF, AIFF, AIF, FLAC, AAC, OGG, OGA, OPUS, M4A, MP4, WEBM, WMA, ALAC, AMRMax size: 50MB.
Or
Split Stems with SAM Audio on Full Songs
SAM Audio is Meta's prompt-based audio separation model. Instead of being limited to fixed stems only, it can isolate sounds described in text prompts, like specific instruments, textures, or effects.
Neural Analog applies Meta's SAM Audio to full songs in stereo, and combines it with specialized models like BS Roformer, MelBand Roformer, and MDX23 for reliable quality.
See SAM Audio in Action
Example: Piano Extraction from Full Mix
SAM Audio extraction using the prompt 'piano' on a complex classical passage.
Example: Removing Crowd Noise from a Live Recording
Effectively remove audience chatter and applause from a live performance, leaving the music intact.
Example: Removing Reverb from a Vocal Track
Clean up vocals by removing room echo and reverb trails using deconvolution.
Example: Removing Drums from a Cinematic Track
Despite the prompt containing NO DRUMS and NO PERCUSSION, the music model still added some. Use SAM Audio to remove these uncommon sounds.
Example: Removing Drums from a Guitar Track
Cleanly separate the drums from a guitar recording to create a drumless backing track.
Examples generated by users
These generations were shared by Neural Analog users
SAM Audio Performance Benchmark vs. State-of-the-Art models (source)
| Model | SAM Audio net win rate | Separation score |
|---|---|---|
| SAM Audio | Reference | 100.0 |
| LALALAI | 8.4% | 91.6 |
| AUDIOSHAKE | 11.4% | 88.6 |
| DEMUCS | 17.6% | 82.4 |
| MOISESAI | 19.5% | 80.5 |
| FADR | 28.3% | 71.7 |
| SPLEETER | 52.8% | 47.2 |
| CLAPSEP | 92.4% | 7.6 |
| FLOWSEP | 93.2% | 6.8 |
| SOLOAUDIO | 95.4% | 4.6 |
| AUDIOSEP | 98.1% | 1.9 |
How to Split Stems with SAM Audio Online
Choose Separation Type
Describe any custom instrument or use a preset to isolate.
Import Your Audio
Drag and drop your audio file or paste a link. We support all major formats.
Download Stems
Our GPU infrastructure processes your audio in seconds. Download individual stems for remixing and mastering.
Where SAM Audio Beats Traditional Stem Splitters
Extract Distorted Voices
Traditional models often struggle to isolate heavily distorted or processed vocals. SAM Audio is notably stronger on these challenging voice separations.
Remove Specific Noise
Do surgical edits by removing one precise SFX class, like barking or laughing, while keeping the rest of the scene intact.
Isolate Same-Family Instruments
Separate lead or solo guitar from rhythm guitar, even when timbres overlap and classic stems are too coarse.
Remove Rare Instruments
Target unusual percussion and niche instruments that standard preset-based models usually miss. SAM Audio is the model to test first for these rare cases.
Restore your stem from mp3 to 20Khz WAV
Stem extraction often contains mp3 artifacts and is limited to 16Khz. Improve the resulting stem using our audio restoration model.
Example: stem of a mp3 music generated with Demucs
Trusted by 30,000+ music lovers
Neural Analog makes me feel like a monkey with an AK-47, in the best way possible
It worked! Well done :) Many thanks :))))
Wow thank you so much i upscale videos and take out live recordings from music because of my autism i hate the crowd
Amazing tool for audio. Clean, simple, and effective. I would spend hours in RX to get the same results. Give it a try. I can save you hours of production time.
I
IMDK
4 out of 5 stars
I love the interface. I love the bulk upload/download features. They’re a life saver! Also I just realized that Apollo is magic and I don’t even need to use denoise. Apollo somehow removes noise much more naturally. So I’m actually spending way less credits than I expected.
TG
The Grim Tower
5 out of 5 stars
Love it! Makes everything crisp!
SH
Shrunken Head
5 out of 5 stars
IT'S GREAT
F
Francisco
5 out of 5 stars
Sensacional
TS
Tommi, Studionet
5 out of 5 stars
Highend services!
B
Bjark
5 out of 5 stars
Easy to use and high quality results.
Listen to tracks made by our users
AI audio enhanced with Neural Analog
Frequently Asked Questions
What is the difference between Neural Analog and the SAM Audio playground?
The public SAM Audio playground is for short mono experiments. Neural Analog puts SAM Audio inside Studio so you can import full songs, preserve stereo workflows, select source regions, and download the resulting stems.
Open full answerStart from SAM Audio Online if you want prompt-based extraction, or use the general stem splitter for standard vocals, drums, bass, and custom prompts.
How long does processing take?
Processing time depends on the operation, selected model, input length, and queue load. Short jobs often finish in about a minute, while long tracks, batch jobs, and heavier models such as SAM Audio Large or UniverSR can take several minutes or more.
Open full answerYou can leave the page and come back later. Finished imports, restorations, masters, and stems stay in Studio, where you can reopen the project and download the result.
How does custom instrument splitting work?
Choose the custom prompt option, then describe the sound you want to isolate, for example "saxophone," "synthesizer," "electric guitar," "crowd noise," or "music with vocals." SAM Audio uses that description to separate the matching layer from the mix.
Open full answerFor difficult sections, open the track in Studio and mark the time span where the target sound is easiest to hear. You can start from SAM Audio Online or the Universal Stem Splitter.
How do I download all stems?
Click the Export button, then choose Download all current stems.
Open full answerThis downloads the stems currently visible in your project, using whatever version is selected for each stem. For example, if your vocal stem is set to a restored version and your bass stem is set to the original version, the export will include those current versions.
To change what gets exported, first select the version you want on each stem waveform, then click Export again.
If you need to create stems first, use the Universal Stem Splitter or open the project in Studio.
Start Using SAM Audio Online
Split stems from full songs, isolate custom instruments, and move faster from generation to release-ready production.