Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 28 additions & 0 deletions catalogue/PROVENANCE.md
Original file line number Diff line number Diff line change
Expand Up @@ -527,6 +527,34 @@ Recraft — V3
$0.04/img raster on every provider (re-verified 2026-09-02; Recraft's
own pricing table agrees).

## Lyria 3 and Lyria 3 Pro — fal serving contracts

The existing `google/lyria-3` and `google/lyria-3-pro` top-level `provider`,
`input` and `output` fields describe their Replicate offering. They are not
universal claims that every provider accepts ten images or emits 48 kHz audio.
The fal bindings retain their independently sourced facts:

- [Lyria 3 API](https://fal.ai/models/fal-ai/lyria3/api) and
[Lyria 3 Pro API](https://fal.ai/models/fal-ai/lyria3/pro/api) require a prompt
and accept one optional `image_url`. Their published input schemas also expose
`enable_safety_checker`; they do not expose separate duration, seed,
output-count or lyrics fields. Tempo, structure and lyrics may be described in
the prompt, which does not guarantee exact control.
- Both API pages describe MP3 at **44.1 kHz / 192 kbps**, rather than the
Replicate-specific 48 kHz fields retained on the card. The standard model is a
30-second clip; Pro is described as up to approximately three minutes. These
output descriptions do not establish caller-selectable durations.
- Both schema pages declare an `audio` File object but show a URL string in their
output examples; the [Pro playground](https://fal.ai/models/fal-ai/lyria3/pro)
also shows a File object. The bindings record this discrepancy without
declaring either example a complete behavioral guarantee. Optional `lyrics`
may contain descriptions and other generated text, not a timed transcript.
- The existing regular fal prices remain unchanged:
[Lyria 3](https://fal.ai/models/fal-ai/lyria3) publishes **$0.04 per audio**,
and [Pro](https://fal.ai/models/fal-ai/lyria3/pro) publishes **$0.08 per audio**.
They are per-output charges, not duration multipliers. Adding these serving
facts does not change model identities, release dates or other providers.

## `pricing: { type: "per_run_flat", usd: 0.04 },`

Source: replicate.com/google/lyria-3 — "$0.04 per output audio file"
Expand Down
70 changes: 68 additions & 2 deletions catalogue/models.json

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

16 changes: 16 additions & 0 deletions models/google/lyria-3-pro/card.toml
Original file line number Diff line number Diff line change
Expand Up @@ -38,9 +38,25 @@ pricing = { "type" = "per_run_flat", "usd" = 0.08, "source_url" = "https://ai.go
[providers.fal]
provider = "fal"
id = "fal-ai/lyria3/pro"
input = "text-or-image"
source_url = "https://fal.ai/models/fal-ai/lyria3/pro"
url = "https://fal.ai/models/fal-ai/lyria3/pro"
pricing = { "type" = "per_run_flat", "usd" = 0.08, "source_url" = "https://fal.ai/models/fal-ai/lyria3/pro" }
qualifications = [
{ source_url = "https://fal.ai/models/fal-ai/lyria3/pro/api", notes = "The request requires a text prompt and accepts one optional image_url. There is no separate duration, seed, output-count or lyrics parameter in the published input schema; lyrics, tempo and structure are described in the prompt." },
{ source_url = "https://fal.ai/models/fal-ai/lyria3/pro/api", notes = "The fal endpoint returns MP3 songs up to approximately three minutes at 44.1 kHz / 192 kbps, with optional generated or supplied lyrics. These serving facts differ from the existing Replicate-specific input and output fields on this card." },
{ source_url = "https://fal.ai/models/fal-ai/lyria3/pro/api", notes = "The output schema declares audio as a File object with a URL, but its JSON example uses a URL string. This documentation discrepancy leaves the exact response representation unresolved." },
]

[[providers.fal.endpoints]]
id = "fal-ai/lyria3/pro"
input = "text"
source_url = "https://fal.ai/models/fal-ai/lyria3/pro/api"

[providers.fal.endpoints.parameters]
prompt = { type = "string" }
image_url = { type = "string" }
enable_safety_checker = { type = "boolean" }

[providers.openrouter]
provider = "openrouter"
Expand Down
16 changes: 16 additions & 0 deletions models/google/lyria-3/card.toml
Original file line number Diff line number Diff line change
Expand Up @@ -38,9 +38,25 @@ pricing = { "type" = "per_run_flat", "usd" = 0.04, "source_url" = "https://ai.go
[providers.fal]
provider = "fal"
id = "fal-ai/lyria3"
input = "text-or-image"
source_url = "https://fal.ai/models/fal-ai/lyria3"
url = "https://fal.ai/models/fal-ai/lyria3"
pricing = { "type" = "per_run_flat", "usd" = 0.04, "source_url" = "https://fal.ai/models/fal-ai/lyria3" }
qualifications = [
{ source_url = "https://fal.ai/models/fal-ai/lyria3/api", notes = "The request requires a text prompt and accepts one optional image_url. There is no separate duration, seed, output-count or lyrics parameter in the published input schema; lyrics, tempo and structure are described in the prompt." },
{ source_url = "https://fal.ai/models/fal-ai/lyria3/api", notes = "The fal endpoint returns 30-second MP3 audio at 44.1 kHz / 192 kbps, with optional generated lyrics. These serving facts differ from the existing Replicate-specific input and output fields on this card." },
{ source_url = "https://fal.ai/models/fal-ai/lyria3/api", notes = "The output schema declares audio as a File object with a URL, but its JSON example uses a URL string. This documentation discrepancy leaves the exact response representation unresolved." },
]

[[providers.fal.endpoints]]
id = "fal-ai/lyria3"
input = "text"
source_url = "https://fal.ai/models/fal-ai/lyria3/api"

[providers.fal.endpoints.parameters]
prompt = { type = "string" }
image_url = { type = "string" }
enable_safety_checker = { type = "boolean" }

[providers.openrouter]
provider = "openrouter"
Expand Down
Loading