A first attempt gives you something to judge. Listen for what the sound suggests, change it for a reason, and try it in the story.
3 examples to hear and remake · 9 briefs awaiting audio and walkthroughs
Set an intention
Make a first take
Listen and judge
Revise with a reason
Compare
Try it in the story
The three examples below were made locally with a small Python recipe, without a paid audio service. Mike reviewed them by listening. The knocks needed a second take; the rain stayed; the radio needed a clearer description. Your judgment may differ.
Remake the three examples on your computer
This is the exact code recipe used for the sounds below. It needs Python 3 installed, but no subscription, account, extra Python packages, or network connection. The walkthrough was checked by reproducing the audio files; a beginner usability trial is still pending.
Open Terminal or your command prompt in that folder.
Run python3 make-effects.py --output my-sounds. On Windows with the Python launcher, use py make-effects.py --output my-sounds.
Open the my-sounds folder. It contains both knock takes, rain, the radio mix, and its three layers as WAV files.
Keep those files. For a variation, edit a copy of the recipe and choose a new output folder. The recipe refuses to replace existing takes.
Prefer recording or an audio editor? Use each brief below as your starting point. Our step-by-step provider and editor walkthroughs are still in preparation; we have not tested those routes for these examples.
Audio + recipe · Listening review: September 21, 2026
Three wooden knocks
Someone arrives just out of view. Make three close, quiet knuckle knocks on a wooden door.
Practice: Rhythm, material, and listening to feedback.
First attempt · too much rattle
Three knocks with a noisy, ringing quality. Retained to hear why the first attempt was revised.
Credit: Created for Renaissance Solutions with Codex using local procedural synthesis; listening direction by Mike Besso. Credit details
What we heard
Mike heard the first attempt as “too snare drummy... or against a metal screen door.”
What changed—and why
We shortened the ringing and confined the noise to the initial contact, using several short, damped tones for the body of the knock. The rhythm stayed the same.
After the revision: “knocks, much better.” Both files remain available so you can judge the change.
How we made it · Three wooden knocks
Local procedural synthesis: short tones plus noise, shaped into impacts.
Make three impacts at 0.10, 0.38, and 0.90 seconds in a 1.5-second clip.
For the first take, mix a decaying 180 Hz tone with a noisy tail. Listen for the drum-like rattle.
For the revision, use short body tones at 145, 335, 570, and 1020 Hz; let each die away quickly. Keep the softer, filtered noise at the point of contact.
Compare at the same peak level. Keep the rhythm fixed while judging the change in material.
Change one thing: Change only the spacing between knocks. Does the visitor seem patient or urgent?
Try it in your story: Place the knocks in a pause before a character reacts. Compare both takes and silence; a technically clean knock can still suggest the wrong door.
Audio + recipe · Listening review: September 21, 2026
Rain against a window
Suggest gentle rain outside a quiet room without pulling attention from the story.
Practice: Background ambience and looping.
First take · selected
Eight seconds of gently varying noise suggesting rain, designed as a repeatable ambience.
Credit: Created for Renaissance Solutions with Codex using local procedural synthesis; listening direction by Mike Besso. Credit details
Change one thing: Change only the delay before the tone arrives. Does the radio feel dependable or hesitant?
Try it in your story: Use it at the moment a vintage radio is switched on. Leave room for the following line of dialogue; do not add it automatically to a modern device.
Nine more sounds to investigate
Each brief has a story purpose and a suggested first experiment. Audio and tested walkthroughs are still in preparation. These are invitations to try, not reports of completed trials.
Workshop brushing and sandingBrief ready · audio to come
Make a 3–5 second cue of careful brushing or sanding at a workbench, gentle enough to sit near speech.
Practice: Texture and repeated action.
Suggested starting path, not yet tested here: Record a brush on unfinished wood, or synthesize soft noise pulses. Keep source takes.
Change one thing: Change only the speed of the repeated action.
Save your first take, note what you hear, and retain any revision with the reason for it. If the first take works, record that. Then listen beside narration and compare with silence.
Two bird chirpsBrief ready · audio to come
Make two short bird-like chirps with a clear rest between them, about two seconds total.
Practice: Pitch, phrasing, and leaving space.
Suggested starting path, not yet tested here: Try a pair of rising synthesized whistles, your own permitted recording, or an existing generative tool.
Change one thing: Change only the gap between chirps.
Save your first take, note what you hear, and retain any revision with the reason for it. If the first take works, record that. Then listen beside narration and compare with silence.
Footsteps on gravelBrief ready · audio to come
Create four walking steps on gravel over 3–5 seconds, with a clear sense of pace and material.
Practice: Pace, material, and distance.
Suggested starting path, not yet tested here: Record safe footsteps on loose gravel; compare a permitted generated take if useful.
Change one thing: Compare calm walking with hurried steps.
Save your first take, note what you hear, and retain any revision with the reason for it. If the first take works, record that. Then listen beside narration and compare with silence.
A door creaking openBrief ready · audio to come
Create a 3-second wooden-door creak whose movement can be matched to a scene.
Practice: Matching sound to movement.
Suggested starting path, not yet tested here: Record a suitable door with permission, or use an existing SFX subscription.
Change one thing: Change only how slowly the door opens.
Save your first take, note what you hear, and retain any revision with the reason for it. If the first take works, record that. Then listen beside narration and compare with silence.
A key turning in a lockBrief ready · audio to come
Make a close, quiet key insertion and turn, with an intelligible final click, about two seconds.
Practice: Small details and precise timing.
Suggested starting path, not yet tested here: Record your own key and lock; preserve separate actions for editing.
Change one thing: Change the space between insertion and turning.
Save your first take, note what you hear, and retain any revision with the reason for it. If the first take works, record that. Then listen beside narration and compare with silence.
Wind through treesBrief ready · audio to come
Create an 8-second ambience of leaves moving in gentle wind, with no speech or music.
Practice: Gradual changes in intensity.
Suggested starting path, not yet tested here: Try a permitted outdoor recording or generation. Inspect wind rumble and any loop seam.
Change one thing: Change only the wind intensity.
Save your first take, note what you hear, and retain any revision with the reason for it. If the first take works, record that. Then listen beside narration and compare with silence.
Distant thunderBrief ready · audio to come
Create a low, distant rumble with a natural fading tail over 5–8 seconds, without a startling close crack.
Practice: Distance and a sound's fading tail.
Suggested starting path, not yet tested here: Try filtered noise synthesis or an existing SFX tool. Do not put anyone outdoors in a storm to record it.
Change one thing: Compare two implied distances.
Save your first take, note what you hear, and retain any revision with the reason for it. If the first take works, record that. Then listen beside narration and compare with silence.
A magical shimmerBrief ready · audio to come
Create a 2–3 second imaginary sparkle that signals a small discovery, not a loud musical sting.
Practice: Designing an imaginary sound.
Suggested starting path, not yet tested here: Layer short synthesized tones, or compare a prompted take.
Change one thing: Change only the pitch direction.
Save your first take, note what you hear, and retain any revision with the reason for it. If the first take works, record that. Then listen beside narration and compare with silence.
A small robot waking upBrief ready · audio to come
Create a 3–4 second sequence of gentle tones and mechanical movement that suggests a small robot waking up.
Practice: Combining tones into a character.
Suggested starting path, not yet tested here: Build from separate tones and clicks, or use an existing SFX tool; retain editable layers.
Change one thing: Change only the rhythm to suggest a different personality.
Save your first take, note what you hear, and retain any revision with the reason for it. If the first take works, record that. Then listen beside narration and compare with silence.
Approach map
Open any approach to inspect its starting material, human work, outputs, and sequence.
Record foley or real objectsCapture a permitted object or performance, then clean and shape it.
Starts with
A cue list · A quiet-enough space · Permission for place and performers
Main actions
Record several takes · Choose and trim · Reduce unwanted noise carefully · Layer and export
Outputs
Original recordings · Edited cue and session record
Human work
Perform the sound · Control the recording · Choose believable takes · Edit timing and tone
Useful when
A tactile, specific sound matters · Original source custody is valuable
Name the cue’s jobDescribe what the audience should understand.
Record safe variationsProtect hearing, equipment, people, and locations.
Edit in contextTrim, clean, layer, and balance against dialogue.
Use a permitted sound librarySelect a reusable source whose license and credit terms fit the publication.
Starts with
A specific cue need · A source with clear terms · A credit and modification record
Main actions
Search by action and material · Read terms for the exact item · Download and preserve metadata · Edit and credit
Outputs
Licensed source file · Edited cue and attribution record
Human work
Judge fit · Verify the item-level license · Keep attribution with the asset
Useful when
A difficult sound already exists · Recording is unsafe or impractical
Search for the actionDescribe material, distance, space, and intensity.
Check the exact itemRecord creator, title, source, license version, and changes.
Edit for the sceneKeep required credit attached through delivery.
Synthesize or generate a cueCreate sound from oscillators, procedural controls, a description, or a mixed sound tool.
Starts with
A precise cue description · Duration and intensity limits · Current terms and export needs
Main actions
Create variations · Reject misleading or noisy results · Edit timing and layers · Record tool, settings, and terms
Outputs
Rendered cue · Prompt, parameter, and edit record
Human work
Direct meaning · Judge artifacts and fit · Confirm rights and disclosure · Finish the audio
Useful when
The cue is abstract or hard to record · Fast variations support exploration
Describe action and acousticsName source, space, distance, sequence, and exclusions.
Choose a creation methodPrefer the control the cue needs.
Synthesis or procedural controlsShape layers: Control attack, body, tail, pitch, and noise.
Description-led generationInspect every variation: Listen for unrelated content, clipping, and unusable tails.
Finish in an editorTrim, balance, label, and export.
Start with your intention
Use My experiment to note your audience, material, limits, equipment, and sharing restrictions. Every field is optional. “Try with our example” fills only an empty goal and constraints field.
Prompt · version 1
Investigate ways to create sound effects
When you know what a moment should communicate but not whether to record, source, synthesize, generate, or layer it.
Map cue-making workflows around a scene job, timing, and rights.
Example: A five-second cue should make a repaired radio feel as if it has powered on.
Show the full prompt
The published prompt stays unchanged.
Source checked ·
Tools to investigate
These five options cover generation, stock recordings, narration, and local editing. The examples above use local procedural synthesis. Provider comparisons are awaiting team trials; source checks do not establish their sound quality. Earlier source dates remain visible in the credits.
This shortlist reflects the date above. Confirm current features, terms, pricing, privacy, exports, and accessibility before choosing.
A high-level workflow
Write a short cue sheet that states the action, perspective, duration, emotional role, required silence, and whether the sound must loop.
Obtain two or three candidates per cue from generation, an approved library, or a local recording; retain the source URL, prompt, settings, license, and raw file.
Edit and level the selected sound locally, audition it against the actual picture or interaction, and export a lossless master plus the required delivery format.
Short custom Foley, ambience, or designed effects from natural-language descriptions when duration and looping controls matter.
How to explore it
Describe one sound or simple event per generation, set duration or looping when needed, compare variations, and download a selected MP3 or WAV for finishing in an editor.
Watch for
Complex sequences are better generated as separate effects and assembled later. ElevenLabs documents paid-plan commercial rights and free-plan noncommercial use with attribution; also review credits and the current output-sublicensing setting before use.
Effects where a novice can describe the sound and optionally perform its rhythm or energy with a voice timing guide.
How to explore it
Use a direct sound description; when timing matters, add a voice performance that marks the desired cadence and intensity, then compare the generated audio against the cue.
Watch for
Voice timing requires microphone access and uploads voice-derived data. Feature labels, account availability, credits, output rights, and whether a beta workflow is suitable for publication need a fresh account-level check.
Finding contributed real-world recordings and unusual ambience when asset-level provenance and Creative Commons filtering are acceptable.
How to explore it
Search for a specific cue, filter by a compatible license, audition the original file, and copy the sound page, creator, title, and exact license into the project's credits record before editing.
Watch for
Licenses differ per asset: CC BY requires attribution and CC BY-NC restricts commercial use. Freesound warns that user uploads can still be unauthorized, and raw samples can trigger automated content claims.
Recording original Foley, generating basic tones or noise, trimming and combining assets, adjusting level, and creating delivery masters locally.
How to explore it
Record or import the source sounds, keep the raw safety copy, edit non-destructively in the project, preview the full cue, and export WAV or the required compressed format.
Watch for
Audacity is an editor and basic generator, not a ready-made sound library or semantic generator. Good results depend on recording technique, editing skill, plugin provenance, and listening checks.
Investigating stock sound effects alongside narration in an existing Studio account.
How to explore it
Check whether your account provides the required stock sounds and usable exports. Keep each source credit with your mix. Our team walkthrough is still in preparation.
Watch for
On September 21, 2026, the Studio page listed stock sound effects and labeled custom AI sound-effect generation coming soon. We have not confirmed or tested custom SFX generation. Studio and the Reader are separate subscriptions.
Use a few criteria and a separate must-have list. An unknown must-have remains unresolved.
Must-have
Story and action fit
The cue communicates the intended event without distracting additions.
Acceptable: A listener can connect the cue to the radio powering on.
Must-have
Timing and clean boundaries
The cue starts, peaks, and ends where the edit needs it.
Acceptable: The five-second file has clean, adjustable boundaries and no clipped tail.
Must-have
Source rights and credit
Every layer keeps creator, source, license, modifications, and credit requirements.
Acceptable: The final cue can ship with a complete attribution record.
Must-have
Listening context
The cue remains useful beside dialogue and on expected devices.
Acceptable: It supports the moment on speakers and headphones without masking speech.
Preference
Layer control and revision effort
Important parts can be changed without rebuilding everything.
Acceptable: The maker retains sources or layers and records corrections, time, and cost.
AI reported
What an AI response says. Follow the sources before relying on it.
Source checked
What an opened source states. This does not prove product performance.
We tested
What RS observed in a dated, bounded trial.
Still uncertain
What is unknown, unverified, or outside the trial.
Prompt · version 1.0
Shape sound-effect criteria one question at a time
Before searching libraries or creating variations, especially when “realistic” is too vague.
Turn an imagined sound into an editable cue brief.
Example: The cue needs a tactile click, a brief wake-up, a warm station tone, and clean five-second boundaries.
Show the full prompt
The published prompt stays unchanged.
Prompt · version 1.0
Compare two cue-making paths in context
After creating two cues from one brief or revising one cue once.
Compare sound cues on meaning, timing, editability, and documented source use.
Example: Compare a recorded and layered radio cue with a synthesized or generated alternative.
Show the full prompt
The published prompt stays unchanged.
Experiment
Make one five-second cue through two feasible paths
Compare a recorded or assembled cue with a synthesized, library, or generated alternative using the same timing and scene job.
Ready-to-use brief
Create a five-second cue for the moment a repaired vintage tabletop radio powers on: a tactile switch click, a brief electrical wake-up, then a warm station tone. It should support the action without sounding dangerous or overpowering dialogue.
Input
Use only sounds you recorded, created, or may reuse. Keep the cue length, scene moment, loudness target, and delivery format fixed.
Limit
Aim for 30–45 minutes per path and no purchase. The local radio example and its recipe are available above. Provider comparisons remain untested.
Path 1
Record or assemble permitted component sounds, then edit and layer them.
Path 2
Create the cue with synthesis, a licensed library, or a permitted generative sound tool, then edit it to the same timing.
Only one candidate? Make one cue, revise only its timing or intensity once, and compare the two versions.
Define success before you begin
Define the cue’s story job, length, loudness context, and success.
Record every source and permission before editing.
Create two feasible paths with the same timing target.
Listen in context on speakers and headphones.
Record edits, retries, noise, export details, effort, cost, and unknowns.
Choose a next step without calling one trial a universal winner.
Observation questions
Does the cue communicate a repaired vintage radio powering on?
Are the start and end clean at the intended edit points?
Does it sit under dialogue at the target loudness?
Can each layer be revised independently?
Are the creator, source, license, modifications, and required credit recorded?
What would you do next? Continue, change an approach, check a claim, try a simpler method, or stop.
Recorded on 2026-09-18
We ran the landscape prompt
Current official-source comparison of ElevenLabs, Adobe Firefly, Freesound, and Audacity for short sound-effect acquisition and finishing.