Audio to session.
Drop in a finished mix and get back a real Logic or Ableton project — every part on its own named track, a MIDI lane under each one, and the tempo the record was actually played at. Not a folder of WAVs.
A finished mix is a sum. Everything that was played is still in there, folded down into two channels — which is why you can hear the bass line perfectly and still not be able to touch it.
Factoring is taking that sum back apart into the terms that made it, and then writing those terms down. Pulling the parts out is separation; working out what was played — the notes, the bar lines, the chords, the tempo the band actually drifted to — is transcription.
Factor does both in one pass, on one grid, and then finishes the job nobody else finishes: it assembles the result into a session you open and start working in.
The same eight bars: the record as it exists, and one player taken out of it.
A six-stem split — vocals, drums, bass, guitar, keys, everything else — is the floor. The trouble with “everything else” is that it is usually a horn section or a string section, which is to say several people, and a stem with five players on it is not a part you can edit.
So when a section is heard, it gets peeled: each player onto their own chair, with their own audio and their own MIDI lane. The summed lane is muted rather than deleted, so nothing plays twice and nothing is lost.
The same happens to guitars when two guitarists were tracked — and a probe decides whether there were one or two before anything is split, rather than splitting on faith.
One stem with five people on it, opened into five chairs.
Most tools hand you one average BPM. That is fine until bar nine, when the take breathes and every region you place after it is early.
Factor writes a per-beat tempo map instead, so the session’s bar lines sit where the band put them, and it makes the first downbeat an actual bar line rather than an offset you discover later.
Then it checks its own work. Every build lints the plan it assembled and reports where each clip sits against the tempo map — so a misalignment is something the build tells you about, not something you find three hours in.
A take that breathes, and bar lines that breathe with it.
What comes back is a native project for the DAW you use — a Logic project or an Ableton Live Set, written directly rather than exported through an interchange format that drops half the arrangement on the way.
Tracks, lanes, markers, tempo, key, pan and colours arrive as project data, which is the difference between a session you can work in and a folder you have to assemble. The chord guide is read from the build’s own transcription, so the chart agrees with the notes underneath it.
The same machinery drives the DAW converter, so a session factored out of a record moves between DAWs afterwards like any other project.
What opens: named tracks, a MIDI lane under each, chords, markers, key.
A mix goes up once and is staged by content, so re-running the same song does not pay for the same work twice. A mix more than 35 cents flat is retuned first.
Six stems on the GPU, then a second peel into per-player chairs wherever a section or a second guitarist is heard.
Beats, downbeats and meter, cross-checked between trackers — a drum-derived grid that disagrees metrically with the master is rejected, not averaged in.
One pass owns every MIDI lane: a lane per stem, a lane per chair, with the stems and the beat grid riding along so notes land on the grid the audio does.
Lay the stems and chairs, build the busses, align everything through one seconds-to-beats map, quantize if asked, then lint the result and report it.
A native Logic project or Ableton Live Set, zipped with its audio, ready to open.
This is the first release under the name, and the number is there because the pipeline is still moving. What v1 fixes in place is the contract: one upload, one grid, one placement path, and a build that reports what it did — which stage took the time, which tracker won the tempo, whether the guitar split actually ran, and whether the final layout lints clean.
What is still being sharpened is named on the roadmap rather than implied by the copy: folding the separate GPU stages behind one client, collapsing four planners into one, and cutting the option surface down to the dials people actually set.
Factor is Doseedo’s audio-to-session pipeline: give it a finished mix, it takes the mix apart into stems and notes, and it hands back a native Logic or Ableton project with everything laid out in time. v1 is the first release under the name.
Separation gives you audio files and leaves the assembly to you. Factor finishes the job: the stems arrive on named tracks, each with a MIDI lane, on a tempo map that follows the performance, with markers, key and a chord guide already in the project.
Logic Pro and Ableton Live, written natively as project data rather than through an interchange format. Other DAWs can open the audio and MIDI that come with it, but the arrangement is what the native formats preserve.
A few minutes for a typical track. Separation, transcription and the per-player splits are each doing real work on a GPU, and the build reports how long every stage took.
Yes — that is the case the extra pass exists for. When a section is heard, each player is pulled onto their own chair with their own audio and MIDI lane, rather than leaving five people summed onto one unusable stem.
Either way. The default keeps the performed timing, because that is what makes a session feel like the record. Ask for quantized and you get notation-ready lanes instead.
That is the normal case, and it is why the tempo map is per-beat rather than a single BPM. The first downbeat is written as a real bar line, and the build lints every region against the map before handing the session over.
There is a free tier to try it on a song. Heavier use and saved sessions need an account; pricing is on the pricing page.