---
type: "post"
id: "earshot-0-4-0-adds-51-recipe-made-d183bf"
slug: "earshot-0-4-0-adds-51-recipe-made-d183bf"
title: "earshot 0.4.0 adds 51 recipe-made sounds, a launch kit preset, and a remade fit engine"
description: "Earshot 0.4.0 adds 51 recipe-made sounds, a music-aware preset for launch films, and a rebuilt sound-fitting engine."
retrieval_nugget: "Earshot 0.4.0 adds 51 recipe-made sounds, a music-aware preset for launch films, and a rebuilt sound-fitting engine. For teams generating audio for product videos, the release offers a more structured starting point, but its own fresh held-out benchmark shows mixed results rather than a clear across-the-board gain."
published_at: "2026-11-12"
updated_at: "2026-11-12"
record_date: "2026-11-12"
date_kind: "published_at"
topics: ["generative-media"]
entities: ["Earshot"]
related_patterns: ["agent-ready-software-exposes-capabilities"]
novelty_label: "new_fact"
what_changed: "Earshot 0.4.1 adds a one-command uv installation path, clearer model and MCP setup guidance, and fixes for non-finite audio samples, media globs, video scans and BLAS thread settings."
prior_refs: []
source_url: "https://raw.githubusercontent.com/Leonxlnx/earshot/main/CHANGELOG.md"
source_urls: ["https://raw.githubusercontent.com/Leonxlnx/earshot/main/CHANGELOG.md"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:earshot-0-4-0-adds-51-recipe-made-d183bf"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/earshot-0-4-0-adds-51-recipe-made-d183bf/","markdown":"https://newruntime.com/posts/earshot-0-4-0-adds-51-recipe-made-d183bf.md","json":"https://newruntime.com/posts/earshot-0-4-0-adds-51-recipe-made-d183bf.json"}
markdown_url: "https://newruntime.com/posts/earshot-0-4-0-adds-51-recipe-made-d183bf.md"
---

# earshot 0.4.0 adds 51 recipe-made sounds, a launch kit preset, and a remade fit engine

## Retrieval answer

Earshot 0.4.0 adds 51 recipe-made sounds, a music-aware preset for launch films, and a rebuilt sound-fitting engine. For teams generating audio for product videos, the release offers a more structured starting point, but its own fresh held-out benchmark shows mixed results rather than a clear across-the-board gain.

## Summary
Earshot 0.4.0 adds 51 recipe-made sounds, a music-aware preset for launch films, and a rebuilt sound-fitting engine. For teams generating audio for product videos, the release offers a more structured starting point, but its own fresh held-out benchmark shows mixed results rather than a clear across-the-board gain.

## What happened
Earshot’s changelog dates version 0.4.0 to October 22, 2026. It describes the release as sound tooling for music-led launch films and says it adds 51 recipe-made sounds across seven new families, bringing the bank to 252 sounds. New categories include cut hits, reverse swells, UI pops, knocks, mallets, drips and shutter sounds; it also adds members to existing categories such as whoosh, riser, swish, glitch and notify. The changelog says the sounds were made from recipes rather than copied from reference clips. Each bank entry carries measured profile and audit data, and the build process rejects sounds that fail its take-cleanliness or originality checks.

The release adds a `launch` kit preset, initialized with `earshot kit init --preset launch`. It selects sounds for common film roles and places them against a music bed. Optional kit-mix settings include the effect level relative to the bed, timing offset against picture, spacing between repeated sounds and stereo width. The preset’s values were measured on the reference clips: effects sit 3–10 dB above the bed, while different sound roles lead or follow picture cuts by specified intervals. Its 72 takes pass the project’s `kit check`. The changelog says the preset is intended for music-led launch films, not that it automates the entire edit or mix.

Earshot also remade its fit engine, which fits sound models to recordings. The changelog describes a new gesture model combining tonal and noise components with motion controls, changes to stereo and room-space fitting, and other updates to timing, beds and spectral detail. It reports a replication benchmark using licensed reference clips that are not included in the repository. The engine was tuned on a fixed 78-clip development split. A 38-clip held-out set was opened once at the end of the first round; a fresh 28-clip held-out set was frozen before further tuning and opened at the end of round C2. No new dependencies were added.

On the fresh 28-clip set, 0.4.0’s median CLAP score was 0.851 versus 0.853 for 0.3, and the mean was 0.860 versus 0.844. The release reports a lower mel gap and improvements in its quality grades on seven clips, with grades falling on two. But only seven takes passed the take check, against eight for 0.3, and onset error increased: 14 of 28 clips were off by more than an octave. The changelog also identifies remaining gaps, including missed pulse counts, clicks in some takes, and bed detection that can choose poorly.

## What the archive adds
New Runtime’s earlier coverage described Earshot as exposing sound description, matching, search and critique as callable capabilities for coding agents ([October 22, 2026 note](https://newruntime.com/posts/earshot-gives-coding-agents-calibrated-hearing-describe-match-1bc321/)). This release is a new capability and benchmark update on that known project, not a restatement of the earlier note.

## Trend signal
This release supports the idea that agent-ready software exposes capabilities: Earshot packages sound selection, rendering and kit checks as operations that software and coding agents can call. The practical boundary is a set of explicit, measurable operations rather than an opaque request to “make the film sound right.” ([Agent-ready software exposes capabilities](https://newruntime.com/patterns/agent-ready-software-exposes-capabilities/))

## Why now
The changelog frames 0.4.0 around music-led launch films and adds a preset whose mix and timing were measured against reference clips. That makes the release timely for builders assembling product-video workflows, while the benchmark cautions against treating the new engine as a uniformly better replacement.

## Why it matters
A sound bank and a preset turn a video-specific task into a repeatable workflow: a team can start from named roles, use sounds designed for those roles, and set their relation to music and picture through explicit parameters. This can reduce manual setup and make results easier to inspect than a one-off generated soundtrack.

The benchmark is equally useful as a warning about evaluation. Aggregate similarity scores do not establish that a sound starts at the right moment or survives a take-level quality check. For production use, teams should test the whole output chain on representative footage and music, including onset, repeated pulses, clicks and bed handling. Earshot’s own results show why a new version number or improved metric should not replace acceptance tests tied to the actual edit.

## Limits and uncertainty
The benchmark uses private licensed clips, so readers cannot inspect or reproduce those exact examples from the repository. The changelog reports aggregate scores and known failure modes, but does not establish how the preset performs across other genres, editing styles or music mixes. It also says the 0.4.0 results preceded a stereo-side fix for hybrid candidates and were not measured again. The project reports no listening test for the new sounds; they were selected and checked by measurement. These figures should therefore be treated as the authors’ evaluation, not independent validation of production quality.

## Skill to build now
If you make product videos, initialize a disposable kit with `earshot kit init --preset launch` and test it against a short edit with your own music bed. Check each role’s timing against the cut, then run `earshot kit check` and listen for clicks, missed pulses and bed artifacts before using the output in a finished film.

## Reader takeaway
Earshot 0.4.0 gives builders a larger sound bank and a music-aware launch-film preset, but its fresh benchmark is mixed. Treat the preset as a structured starting point, and validate timing and take quality on the footage and music your team actually ships.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Generative Media](https://newruntime.com/topics/generative-media.md): Explore the generative-media topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)