Custom AI Sound Effects from Text Guide
Step-by-step instructions for generating custom sound effects from text prompts inside Flixly using specific models and credit workflows.
TL;DR
Flixly produces custom AI sound effects from text via Gemini 3.1 Flash TTS for short clips and Seedance 2.0 for longer tracks. Control timing with explicit tags, pay 2-4 credits per 10 seconds, and export at 44.1 or 48 kHz. Follow the eight-step workflow for repeatable results.
Field overview
Roughly 12 major platforms offer text-to-audio conversion today. The axis that actually separates them is output control over duration, sample rate, and prompt adherence rather than raw model size.
Landscape of text to audio options
Flixly supports sound effect generation through its music and speech tools. Users access Gemini 3.1 Flash TTS at /dashboard/text-to-speech for short clips under 15 seconds. Music Generation at /dashboard/music-generation handles longer atmospheric tracks up to 90 seconds when prompted for effects.
Other entries include standalone services that limit free tiers to 5 exports daily. Flixly instead charges credits per generation, with a base of 2 credits for a 10-second clip at 44.1 kHz.
Dimension that matters most
Prompt precision determines final quality. Models that accept explicit timing tags and style descriptors produce fewer artifacts than those limited to plain sentences.
Flixly lets you specify start and end times inside the prompt field. A sample input reads "door creak slow 0-3s then slam at 4s" and returns a 5-second file at 48 kHz.
Head-to-head comparison
Seedance 2.0 integrated via the music tool yields tighter timing than basic TTS. Veo 3.1 handles layered effects when used in video-to-audio pipelines. Kling 3.0 offers higher sample rates but consumes 4 credits per 20-second output.
A quick reference table follows.
| Model | Max duration | Sample rate | Credits per 10s | Prompt tags supported |
|---|---|---|---|---|
| Gemini 3.1 Flash TTS | 15s | 44.1 kHz | 2 | time, intensity |
| Seedance 2.0 | 90s | 48 kHz | 3 | style, fade |
| Kling 3.0 | 60s | 96 kHz | 4 | layer, reverb |
Use case picks
Pick /dashboard/text-to-speech if your clip needs under 15 seconds and exact word timing. Pick /dashboard/music-generation when atmospheric layers or 30-plus second builds are required.
Step-by-step generation process
- Log in and navigate to the dashboard. Select the text-to-speech tool first.
- Enter a prompt under 80 characters that names the sound and timing markers.
- Choose Gemini 3.1 Flash TTS from the model dropdown.
- Set output length to the nearest whole second and sample rate to 44.1 kHz.
- Generate the preview and review waveform for clipping.
- Adjust prompt tags if the initial file has gaps longer than 200 ms.
- Export the final WAV or MP3 file once credit balance allows.
- Store the file in your project folder with a descriptive name including the model and duration.
Credit costs and limits
A single 10-second effect at standard quality uses 2 credits. Bulk runs of 50 clips require a minimum balance of 120 credits purchased via the pricing page at /#pricing.
Users report average sessions consume 18 credits when iterating on the same effect three times.
Inline tool references
Start with the Text to Speech page for quick tests. Move to Music Generation for complex layers. Combine outputs in Video to Video if you later need synced footage. Reference the Voice Cloning tool when effects require human-like textures. Finally, compare against external options via the alternatives page for ElevenLabs.
Limitations to expect
Prompts longer than 120 characters often truncate. Outputs above 48 kHz require the pro credit tier. No native support exists for spatial audio formats such as ambisonics.
The platform does not generate files larger than 10 MB per request.