Vortex Creative Center · 26 October 2026 · pull request #5, waiting for your go
Making a video, revised after your first real one
You drafted Street Booth on the live studio and hit five problems. All 8 voice lines failed with the provider's error, shots took dropdowns to inspect, setting up asked for things the recipe already knows, there was no picture-and-text stage before video, and the characters didn't speak. Each one is fixed below. The screenshots come from a private test copy (grey test pictures, nothing spent).
PR #5 · open60 of 60 tests passNot live yetNo money spent
What changed
Storyboard first, one panel per shot, characters speak
1The voice error is gone from your sightKIE’s own record: a 500 on their side, 0 credits charged. Now a piece that fails on the provider’s side is sent again quietly (up to 4 times), and a voice line moves to ElevenLabs Multilingual v2 after two tries. If something really can’t be made, the card says why in plain words, with “Try again” and its price.
2Click a shot, see everythingOne panel: the picture and its takes; who says the line and in which voice; the start frame with what it’s made from and its words (edit them, then New start frame · $0.04); how the clip is made; what the rulebook asks of that part; notes and history.
3Three picks to startA recipe, our game, a market. Everything else is filled in from the recipe and the market, folded away, and changeable later shot by shot.
4Storyboard first, then video“Make the storyboard” (≈ $0.45): the cast & set, a start frame per shot and the end card, with who says what under every card and the full script. Video only after you’ve looked: “Make the clips”.
5Characters speak, in their own voicesEach shot now knows who says its line and whether we see their mouth. The four Street Booth shots where the host or the passer-by speaks on camera are made from the start frame + that character’s voice line → lipsync (InfiniteTalk). Two people, two voices: the host is James, the passer-by Rachel M, the end card a narrator. Her question from behind stays a voice-over.
The screens
Tap a picture to see it whole.
01The storyboard, and the scriptPictures and who says what first; no video until you say so.+
Videoafter “Make the storyboard”
The cast & set and every start frame are in (grey test pictures). Each card says who says its line and how: “Booth host · says it on camera”, “Passer-by · voice-over”. Next: Make the clips.
Videothe script
Who says what, shot by shot, with each voice: Passer-by (Rachel M), Booth host (James), the narrator (Cassidy) on the end card. “Make the clips · $0.48”: 2 clips from their frames, 4 speaking on camera, lipsynced, 8 lines in their speakers’ voices, the music.
02One panel per shotEverything about the shot in one place: its line and voice, its start frame and words, its clip, the rulebook, notes.+
Shot 2Host answers
Who says what: the booth host, on camera, so the clip is lipsynced to the line. Change the line, or his voice in every shot he speaks.
Shot 2its start frame and clip
Start frame, made from the booth host, the passer-by, the booth and the game’s logo, with the words it shows. New start frame · $0.04, or use your own picture. The clip: the start frame says the line in James’s voice, lipsynced (InfiniteTalk, $0.06). What the rulebook asks: “only his face needs lip sync”.
Shot 2on a phone
The same panel on a phone.
New videothree picks
Recipe, game, market: “That is all it needs”. Set up the video; the storyboard next (≈ $0.45). The details are folded below, filled in.
What I need from you
Say “merge and deploy”This puts PR #5 and the merged PR #4 (the pipeline, Our games) on the live studio, with a database backup first. Your Street Booth video then shows “Try again” for its 8 voice lines, and they come back in the host’s and the passer-by’s own voices.
Then make one Street Booth for real“Make the storyboard” (≈ $0.45), look at it, then “Make the clips”. The real lipsync quality on a two-person frame is the thing I haven’t seen yet.
Real output. No paid run. The retries, the backup voice model, the lipsync requests and the voices were all checked with test providers, including a test where KIE’s first voice model keeps failing.
Lipsync with two people in the frame. InfiniteTalk is made for one speaker. In Street Booth only the speaker’s face is seen in shots 2, 5 and 6, but the passer-by’s shot 7 also shows the host, so check that one first.
Other languages. The lines are still English for every market. The voices are English voices.
