When Foundry launched, it was a music generation studio. You typed a description, it produced a full song with vocals and instruments, and then you could edit, layer, and mix the result. That part is still there and stronger than ever.
What changed: Foundry now generates speech too.
What speech generation looks like in practice
You write or paste a complete script into the rich Speech Editor. You assign voices to roles and direct the performance where it changes. Speakers, delivery styles, music, effects, delays, overlaps, and crosstalk all stay inside the same readable document.
Foundry has 38 delivery styles across emotion, drama, narration, and character states, each with 5 intensity levels. You can direct joy, anger, heroic calm, documentary narration, exhaustion, whispering, singing, and much more. The voice stays consistent across the entire read without flattening every performance into the same neutral tone.
60 built-in speaker presets cover a wide range of vocal types. If none of them match what you need, voice cloning lets you create a custom voice from a short audio sample. That cloned voice then works with the same delivery controls as everything else.
Multi-speaker scenes
Audiobooks have multiple characters. Podcasts have co-hosts. Dialogues need distinct voices that interact naturally. Foundry handles all of this in one document. Speakers can alternate, overlap, interrupt each other, or speak in natural crosstalk while each voice stays separate and recognizable.
A full audiobook chapter with three characters, each with their own emotional arc, generates at up to 15x real-time speed. On a 12 GB NVIDIA card, that means a 10-minute scene is ready in under a minute.
Music and speech in the same narration document
This is where the combination gets interesting. You can place background music, effects, and exact timing directly around the narration. Preview, Play All, and export use the same sample-accurate arrangement, so the finished result matches what you heard while editing.
When a production needs deeper timeline work, you can still send a speaker, paragraph, or complete document to Tracks. Normal narration no longer requires manually arranging spoken clips there.
Podcast intros with custom music beds. Game trailers with character dialogue over an original score. Audio dramas with layered sound design. All built inside one application, all running on your hardware.
Everything else is still here
Text-to-music generation with the caption builder. Patch editing to fix sections without starting over. Stem separation into 7 tracks. Cover and restyle workflows. 32 base DSP effects with hundreds of presets, sub-effects, and audio modifications that can be combined. The Creative AI for writing assistance. The full timeline editor with multi-track mixing.
Speech did not replace any of that. It just made Foundry into something broader: a complete local audio production environment for both music and voice.