AI SFX Lab vs AudioGen

AI SFX Lab is a web tool for generating custom sound effects from text prompts. Users can set duration, ask for variations, choose loop behavior and style, and specify sounds the AI should avoid. A prompt enhancer can refine a description before generation. The maker says results may differ between generations and recommends previewing clips, changing settings, and trying again. The free plan costs nothing and provides 100 credits each month, with generations up to five seconds; it has no downloads or commercial use. Starter costs $4.90 a year, billed annually at $59/year, and includes 600 monthly credits. Creator costs $9.90 a year, billed annually at $119/year, with 2,000 credits and up to three simultaneous generations. Pro costs $24.90 a year, billed annually at $299/year, with 7,000 credits and up to five simultaneous generations. These paid plans support clips up to 22 seconds and downloads in WAV, MP3, and FLAC. Monthly credits do not roll over. Commercial use is restricted to paid plans and governed by the maker’s terms and policies, which prohibit deceptive audio, rights infringement, and misleading emergency sounds. The product is aimed at creators, game developers, editors, podcasters, streamers, marketers, Foley artists, and creative teams.

AudioGen is a text-to-sound model provided through AudioCraft. It generates audio samples from text descriptions and can continue from an audio prompt, with conditional and unconditional generation modes. Sampling controls include greedy, temperature, top-K, and top-P methods. The listed pretrained model, facebook/audiogen-medium, has 1.5 billion parameters; the example API generates sound and saves it as WAV. The implementation uses an autoregressive Transformer and a 16 kHz EnCodec tokenizer. The repository provides API access and a local Jupyter notebook demo. Running the medium model locally requires a GPU with at least 16 GB of memory; installation requires Python 3.9 and PyTorch 2.1.0, with ffmpeg recommended. The model was trained on English descriptions and performs less well in other languages. It cannot generate realistic vocals, and prompts may need refinement. Code is MIT-licensed, while weights use CC-BY-NC 4.0; training datasets are not provided. The listed self-hosted plan is free.