Pel&Bee Lab

ScoreFlowPrivate beta

Spots, selects and edits library music to picture.

ScoreFlow is our own DAW plus a team of Claude agents that does the job of a TV music department. Used in-house on our broadcast TV projects.

Request beta accessRead the lab notes
The ScoreFlow DAW: an episode timeline with scene boundaries, music cue regions labelled by family, dialogue tracks and music clips placed on the scoring tracks.
ScoreFlow DAW — music regions placed by the agents

Pipeline

Five steps and a learning loop.

Each step is a Claude agent with one job. The composer stays in charge of what ships.

  1. 01

    Picture

    Cut detection, faces, voices and transcripts, fused into a character layer.

    Scene analysis
  2. 02

    Scene book

    Claude agents write a per-scene book of the episode in parallel, then check it and freeze it.

    Parallel agents
  3. 03

    Spotting draft

    Places music regions from the book, each with one of 20 fixed music families (love theme, tension, comedy, hope, action…). Entries snap to picture cuts; target music share is about 70%.

    20 families · ~70%
  4. 04

    Track selection

    Picks tracks from the library by family and energy, with no repeat inside 10 minutes. Edits join on cuts and scene boundaries; bar jump-backs fit the length.

    Family · energy
  5. 05

    ScoreFlow DAW

    The agents place everything on the music tracks; the composer reviews, edits and delivers.

    Composer review
06

Learning loopAfter every delivery, the system compares its picks with the composer's final mix (kept / swapped / silenced) and learns from the composer's edits.

Spotting draft

Illustration of a spotting draft: picture cuts, scene changes, dialogue and music regions with their families, and the in and out of one cue.

Measured on our own deliveries

  • Cues kept88/100

    of the last 100 cues it placed were kept in the final mix

  • Music / silence0.82

    agreement with the composer's final mixes

  • Scene changes~7in 10

    found, and right 91% of the time when it marks one (F1 0.80, 21 episodes)

  • Full run34.7min

    for a 35-minute episode

Cues kept in the composer's final mix, episode by episodeFrom 38% to 87% in 13 episodes.
  • Kept, per episode
  • 3-episode rolling average

Share of music regions the system placed that the composer kept unchanged in the delivered mix. In-house, one 90-episode broadcast drama, Sept–Oct 2026.

Same melody repeated within 10 minutes, per episodeFewer repeats as track selection learned the episode.

Count of times the same melody came back within 10 minutes in an episode's music. In-house, same broadcast drama, Sept–Oct 2026.

Shipping daily

Shipping daily.

ScoreFlow gets measurably better every week. Recent steps are in the lab log.

One 35-minute episode on our GPU workerA full run now finishes faster than the episode plays.
  • Then
  • Now

35-minute broadcast episode, Oct 2026, in-house runs. Music regions land 28.2 minutes sooner.

The DAW

The DAW: local-first, cloud when you want it

Think Dropbox, but for a DAW.

  • Local-first

    Works on your own drives, offline, at full speed.

  • Cloud sync on demand

    Hit sync and the project goes up; otherwise everything stays local.

  • Streaming-ready

    Video, dialogue, stems and WAVs are converted for fast streaming on upload.

  • Open anywhere

    At a post house for revisions, open the project from a private web link. No install.

  • Revise and deliver online

    Make changes in the browser and export, or download the project.

  • Import on the go

    New material from the studio is imported and read offline too.

Next

Original music, with no generative audio.

Claude writes the MIDI, licensed instrument libraries play every note in the DAW, and the system bounces its own mix. Nothing is generated as audio, so the rights stay clean for broadcasters.

Try ScoreFlow on one episode.

Private beta for broadcasters and studios.

Request beta access