Audio for video · Sound & text-to-audio
AI Sound Effect Generator
Write a cue-sheet prompt, choose Background sound or Transitions & hits, and download MP3 to stack under voiceover or AI video.
- Text-to-audio & ambience
- Seed Audio 1.0 via Fal
- Layer under VO & AI clips
Try the generator
Choose what you need + a preset — then edit the cue before you generate.
What do you need?
Start from a preset
Target length
Seed Audio follows your description — we add a length hint to the prompt (no separate duration API yet).
Seed Audio 1.0 via fal.ai — text-to-audio for backgrounds, transitions, and mixed scenes. Import MP3 as a separate track in your edit.
Your audio
Your audio appears here — generate to preview and download.
12 credits (estimate)
Targeting ~30s output
Layer audio in your edit
Generate beds and stings as separate MP3 layers — align transients to cuts, keep Shorts mixes minimal.
Step 1
Write cue
Bracket ambience and impacts. Say what should NOT appear (e.g. no vocals) when you only need SFX.
Step 2
Pick layer
Place under voiceover — duck music when VO plays; nudge whooshes to frame transitions.
Step 3
Download MP3
Credits shown upfront. Preview length and tone before download.
Step 4
Mix & export
Export AI video from Reeldo with sound-on models, or mute native audio and use your custom bed.
Sound scenarios for video
Chips pre-fill prompts tuned for common creator jobs.
Ambience bed
Loop-friendly room tone for talking-head B-roll or desk scenes.
Use in generator →Impact / whoosh
Short transients synced to title cards and jump cuts.
Use in generator →Trailer stinger
Epic hits for launch montages without licensing stock libraries.
Use in generator →Podcast open
Intro music and spoken welcome in one text-to-audio pass.
Use in generator →AI video bed
Neutral pulse under faceless tutorials when model native audio is too busy.
Use in generator →
Example sound jobs
Listen to samples for each job type, then generate your own cue.
Sample
Sample
Sample
Sample
Voiceover vs sound design
Use voiceover when you have a script to read. Use text-to-audio when you need ambience, stings, or mixed scene beds for video.
| Topic | AI Voiceover | AI Sound / text-to-audio |
|---|---|---|
| Input | Finished script (characters) | Director-style prompt (cue sheet) |
| Output | Single clean narration MP3 | Mixed scene audio (SFX, beds, lines) |
| Best for | Shorts hooks, courses, ads, faceless VO | Ambience, whooshes, podcast opens, AI video beds |
Sound & text-to-audio troubleshooting
Seed Audio follows your prompt — length and density are controlled with target length, layer choice, and negative constraints.
Clip shorter or longer than target
Target length is a prompt hint — not a hard API duration.
Regenerate with tighter wording, pick a closer Target length chip, or trim silence in the NLE.
Bed feels crowded or noisy
Mixed scene prompt or too many adjectives.
Switch to Background sound, add “no drums, no vocals, minimal motion”, and use ~30s background tier.
Unwanted singing or speech
Full scene mix or lyrics in the prompt.
Use Background sound or Transitions & hits and state “no spoken words, no lyrics” explicitly.
Bed drowns voiceover
Bed at full volume under narration.
Lower bed −12 to −18 dB, high-pass around 120 Hz, sidechain duck when VO plays.
Whoosh doesn’t land on the cut
Transient not aligned in the timeline.
Generate a Quick hit (~3s), import to the cut point, nudge by frames — don’t stretch the MP3.
Background sound cue
Soft rain on glass, distant city hum, calm desk mood. No vocals, no drums, loop-friendly under narration.
Transition hit cue
Cinematic whoosh into a single sub hit for a title card. Clean tail, no music bed, no speech.
Mixing & prompt tips
Tip 1
Duck beds under VO
Keep one VO track master — automate bed gain down 6 dB whenever narration plays.
Tip 2
Negative prompts matter
List what to exclude (vocals, EDM, crowd chant) — especially for beds and impacts.
Tip 3
Check on phone speakers
Shorts mixes should leave headroom; one bed OR one hit per hook is usually enough.
Sound credits by target length
Credits map to the Target length you pick before generating. The bar shows the exact cost upfront.
| Target length | Credits | Typical use |
|---|---|---|
| Quick hit (~3s) | 8 | Whoosh, UI tick, cut accent |
| Transition (~8s) | 8 | Stinger, logo reveal tail |
| Background (~30s) | 12 | Room tone, tutorial bed |
| Long scene (~60s) | 18 | Podcast open, longer ambience |
Prompt limit 2,048 characters. Output length varies with Seed Audio — use preview and trim in your editor.
Sound under AI-generated video
Generate clips in Studio, mute or reduce native audio, and place your Seed bed + voiceover in the same timeline.
