Fish Audio
Text to speech in 80+ languages, and voice design.
Tools1
media__generatePer modelGenerate an image, video or audio clip and get back signed URLs to the stored output. Required: `model` (from media__list_models) and `input` (per media__get_model's input_schema). File-typed inputs accept a public https URL or a Drive path / file id (an upload via drive__create_upload, a connector download, a mail attachment, an earlier output). Waits inline up to `wait_seconds` (default 45, counted against the WHOLE call; images usually finish well inside it). The server responds within the budget (finishing files right at the buzzer may add a few seconds), so prefer NO client-side timeout for this call — or a backstop of at least 2× wait_seconds. A longer generation returns `{job_id, status: "running"}` and KEEPS RUNNING server-side — collect it any time with media__get_result, from any session. If your client times out mid-call, nothing is lost: the newest entry in media__list_jobs is your job — do NOT resubmit. For expensive kinds (video), check the price with media__get_model first. Cost: a successful generation is billed at the model's rate times the usage the upstream actually reports, so the final amount can differ from the rate card for calls whose size is only known after the fact. Failures, cancels and expiries never charge.Generate an image, video or audio clip and get back signed URLs to the stored output. Required: `model` (from media__list_models) and `input` (per media__get_model's input_schema). File-typed inputs accept a public https URL or a Drive path / file id (an upload via drive__create_upload, a connector download, a mail attachment, an earlier output). Waits inline up to `wait_seconds` (default 45, counted against the WHOLE call; images usually finish well inside it). The server responds within the budget (finishing files right at the buzzer may add a few seconds), so prefer NO client-side timeout for this call — or a backstop of at least 2× wait_seconds. A longer generation returns `{job_id, status: "running"}` and KEEPS RUNNING server-side — collect it any time with media__get_result, from any session. If your client times out mid-call, nothing is lost: the newest entry in media__list_jobs is your job — do NOT resubmit. For expensive kinds (video), check the price with media__get_model first. Cost: a successful generation is billed at the model's rate times the usage the upstream actually reports, so the final amount can differ from the rate card for calls whose size is only known after the fact. Failures, cancels and expiries never charge.
Models4
- Fish Audio Drama 315 credits / 1000 UTF-8 bytes
- Fish Audio S2.1 Pro22 credits / 1000 UTF-8 bytes
- Fish Audio S2 Pro22 credits / 1000 UTF-8 bytes
- Fish Audio Voice Design15 credits / request