Audio models
| Model | Input | Does | Cost |
|---|---|---|---|
seed_audio |
Text or Audio | Speech and sound effects | 0.25 credits/second (5 credit minimum) |
eleven_v3 |
Text | Text to speech | 1 credit per 50 characters (1 credit minimum) |
eleven_multilingual_v2 |
Text | Multilingual text to speech | 1 credit per 50 characters |
eleven_text_to_sound_v2 |
Text | Sound effects | 1 credit/second, or 2 credits with no duration given |
eleven_voice_isolation |
Audio | Strips background from a voice track | 1 credit per 6s |
eleven_voice_dubbing |
Audio | Dubs into another language | 1 credit per 2s |
eleven_multilingual_sts_v2 |
Audio | Speech-to-speech conversion | 1 credit per 3s |
Picking one#
Narration and dialogue. eleven_v3 is the current-generation text-to-speech model; eleven_multilingual_v2 when you need many languages from one voice. Both bill per 50 characters, so a 500-character paragraph is 10 credits — ten cents.
Sound effects. eleven_text_to_sound_v2 generates a sound from a description. Supplying a duration bills 1 credit per second; omitting it bills a flat 2 credits, which is usually cheaper for short effects.
Cleaning up recordings. eleven_voice_isolation removes background noise from a voice track at 1 credit per 6 seconds — a ten-minute recording is 100 credits.
Localising existing audio. eleven_voice_dubbing is the most expensive transformation at 1 credit per 2 seconds, because it transcribes, translates and re-synthesises.
Changing a voice without changing the words. eleven_multilingual_sts_v2 maps one performance onto another voice at 1 credit per 3 seconds.
Runway's own model. seed_audio handles both text-to-speech and sound effects at 0.25 credits per second — the cheapest per-second rate here — with a 5-credit minimum per generation.
Supported audio formats#
Models that accept audio take these containers and codecs:
| Container | Extension | Content type | Codecs |
|---|---|---|---|
| MP3 | .mp3 |
audio/mpeg, audio/mp3 |
MP3 (MPEG-1/2 Layer 3) |
| WAV | .wav |
audio/wav, audio/wave, audio/x-wav |
PCM (uncompressed) |
| FLAC | .flac |
audio/flac, audio/x-flac |
FLAC (lossless) |
| M4A | .m4a |
audio/mp4, audio/x-m4a |
AAC, ALAC |
| AAC | .aac |
audio/aac, audio/x-aac |
AAC (raw) |
Audio inputs are capped at 32MB by URL and 16MB as a data URI. Above that, use ephemeral uploads, which take up to 200MB.
The Content-Type header must match the actual media type — file extensions in URLs are ignored, and application/octet-stream is explicitly rejected.
Related#
- Image and audio pricing — full cost tables
- Inputs — size limits and header requirements
- Video models —
veo3.1generates video with native audio
A 500-character narration costs 10 credits — ten cents.