feat(cli): add video-to-video-music and video-to-video-sfx - #18
Merged
Conversation
The API has 11 generation endpoints and the CLI covered 9: both
video-to-video endpoints had SDK resources and public docs, but no
`sonilo` subcommand. Both are async-only and return the source picture
with the generated audio muxed in, so they follow video-to-video-sound:
one file out, default `output.mp4`, no --format.
video-to-video-music takes --prompt, --preserve-speech and
--isolate-vocals; unset switches forward None rather than false so the
server default stands. It takes no --segments, matching the resource.
video-to-video-sfx takes --prompt and --segments in the SFX shape
{start, end, prompt}, reusing the existing _add_segments helper and the
shared shape-checking reader, so a music-shaped array is rejected before
any request is made.
Bumps sonilo-cli to 0.5.0 in pyproject.toml and __init__.py together,
since the latter feeds the x-sonilo-client-version header.
Both endpoints OR the two fields into a single behaviour server-side — video_to_video.py does `keep_speech = bool(preserve_speech) or bool(isolate_vocals)` and video_to_music.py does the same normalization onto its own local — so --isolate-vocals is the legacy name for --preserve-speech, not a second feature. The help text on both commands claimed it split out a vocals-only stem. On video-to-video-music that describes something the endpoint cannot do at all: the result is one muxed video with no stems. On video-to-music the flag was pre-existing and equally misleading, since it presented two names for one feature as two independent options. Also corrects the two matching claims in the CLI README and adds a test so neither help string can drift back.
Sapient docs evalsWaiting for the staging docs URL before running evals. Sapient will start the selected PR evals automatically when GitHub reports a successful docs deployment for this PR. This usually happens within 15 minutes. Commit: |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The API ships 11 generation endpoints; the CLI covered 9.
video-to-video-musicandvideo-to-video-sfxhad SDK resources and public documentation but were never wired up as CLI subcommands. This closes that gap, so the CLI now covers every generation endpoint.The two commands
Both are async-only and write a video, modelled on the existing
video-to-video-sound:video-to-video-music— exactly one of--video/--video-url, plus optional--prompt,--preserve-speech,--isolate-vocals,--output(defaults to an.mp4)video-to-video-sfx— the same, with--segmentstaking SFX-shaped{start, end, prompt}items, reusing the existing segments reader and shape rather than a second copyvideo-to-video-musicdeliberately has no--segments: the request builder cannot send segments to that endpoint at all, and a test pins that.Cross-CLI parity
The Node and Python CLIs were implemented against one shared contract and their help output was compared directly afterwards. The flag surfaces match exactly.
A pre-existing bug fixed along the way
That comparison surfaced a long-standing error on the existing
video-to-musiccommand in both CLIs. Its help presented--isolate-vocalsand--preserve-speechas two independent options with different descriptions, but the backend ORs them into a single behaviour —backend/app/routers/v1/video_to_music.pynormalises the two under a comment namingpreserve_speechthe current public name andisolate_vocalsthe legacy field name. One feature was being documented as two.Both CLIs now describe
--isolate-vocalsas a legacy alias, with--preserve-speechlisted first, and a test pins the wording so it cannot silently regress. The endpoint does produce vocals and mux tracks, but neither CLI can retrieve them — there is no flag to fetch anything but the main audio — so advertising a stem would trade one misleading line for another.Tests
83 tests pass in
sonilo-cli(15 new for the two commands, plus a parametrised guard on the--isolate-vocalswording). Root suite 192 pass andsonilo-video-kit53 pass, both untouched.The result is saved through the SDK's own
VideoResult.save()rather than reaching into the result's fields, so the CLI cannot drift from how the SDK models the artifact.Release
sonilo-cligoes 0.4.0 → 0.5.0 in bothpyproject.tomland__init__.py— the latter feeds thex-sonilo-client-versionheader, so the two move together. The coresonilopackage is not bumped.