Ask LTX or Grok to continue a keeper, not a clip you still dislike
extend_video generates new frames after the last moment. Continuing a miss, or extending before picture lock, pays per second of a sequel you will not cut.
3 min read
Replace one slide after the carousel order is the one you will render
replace_slideshow_slide bills a new image; remove and reorder do not refund or generate. Surgery on a board you have not locked, or on a reel you already exported, is the expensive order.
3 min read
Grab a thumbnail frame from the cut you will actually post
extract_video_frames at one timestamp is a poster from existing pixels. Pulling a thumb off a miss, or before picture lock, publishes a still of a take you will replace.
3 min read
Transcribe the keeper, not the miss you are about to replace
transcribe_audio (Cartesia ink-whisper) returns plain text. Running it on a throwaway soundtrack, or before picture lock, pays for a script of words you will never reuse.
3 min read
Isolate the voice you will ship, not the noisy take you will regen
isolate_audio pulls a vocal out of a mixed clip. Isolating a miss, or isolating before picture lock, spends a rescue on audio you will never mix.
3 min read
Pin a brand kit after you know the look, not while tasting presets
set_brand_kit stores caption style, fonts, and colors as defaults. Pinning a DYNAMIC look too early doubles the caption bill on every later clip, including misses.
4 min read
Animate a portrait to speech after the still and the line are locked
generate_lipsync needs a face image and an audio file. Driving a rejected still, or a scratch read, spends a lipsync model on inputs you will replace.
3 min read
Crop to 9:16 after you know which take survives review
edit_video at 9:16 is the TikTok/Reels/Shorts wrap. Verticalizing a miss, or cropping before picture lock, throws away side pixels you paid to generate and bills an export you will redo.
3 min read
Stitch scenes after each clip is a keeper, not as a way to hide misses
merge_movie_scenes concatenates clips with a shared transition. Merging a pile of maybes, or merging before picture lock, makes a sequence you will recut clip by clip.
3 min read
Read insights before you reframe a catalogue of throwaways
get_content_insights recommends a format from your own posts. The credit leak is running edit_video on unlocked clips in the direction those insights point.
3 min read
Cut out the still you will actually animate, not the rejected board
Photo cutouts go through generate_image_from_image (and sometimes remove_background). Erasing a rejected still, or erasing before the board is locked, spends an image edit on a tile you will not use.
3 min read
Key a black-matte keeper, not a talking head you will regenerate
remove_video_background is FFmpeg colorkey for pure-black backdrops; remove_background is the model path. Keying a miss, or keying before picture lock, keys a ghost.
3 min read
Strip-and-replace audio only after picture is allowed to freeze
attach_audio_to_video in replace mode mutes the original and lays a new track. Doing that on a miss, or before picture lock, destroys a soundtrack you might have needed and bills a swap you will repeat.
3 min read
TikTok reshape and trim wait until the source cut is locked
edit_video plus cut_video make a 9:16, scroll-length file. Doing both on a miss, or before picture lock, is a TikTok export of a take you will not post.
3 min read
Snapshot an EDL after the structure is the structure, not while tasting clips
save_editor_project stores timeline state. Snapshotting a pile of maybes, or saving before picture lock, freezes a draft that will teach next week's edit the wrong footage.
3 min read
Slice a variation grid after you know which tiles are worth upscaling
split_image_grid is a deterministic cut on the rows and columns you name. Slicing a sheet you will discard, or slicing before you pick winners, creates tiles you will still have to throw away — often after an upscale.
3 min read
Split stems after the generated track is the bed you will mix
separate_music_vocals needs a Suno taskId and audioId. Splitting a discarded generate, or splitting before picture lock, yields stems for a song you will not attach.
3 min read
Audio-only translation should wait for the master soundtrack
dub_video with translate_audio_only is a lighter language pass. It still bills against length, still runs async, and still dies if you recut the parent.
3 min read
Trim the keeper's in and out, not a take still in taste
cut_video with one KEEP range is a clean trim. Trimming a miss, or trimming before picture lock, produces a short leftover you will not ship.
3 min read
Render a slideshow as a reel only after overlay text is signed
convert_slideshow_to_video is a render pass over slides you already paid for. Converting a draft board, or converting before overlays and order are locked, makes an MP4 you will have to re-render.
3 min read
Pull frames from a locked clip, not from a miss you will overwrite
extract_video_frames returns stills from a video. Grabbing a contact sheet off a throwaway take, or extracting before picture lock, produces photos of a performance you will replace.
2 min read
Sharpen the still you will publish, not every tile in the grid
upscale_image enhances existing detail. Upscaling rejected stills, or upscaling before the board is locked, is a second generate you will throw out.
3 min read
Approve lyrics against a locked runtime, not a throwaway song
generate_lyrics returns text. generate_music then spends a track. Writing lyrics before picture lock, or scoring a miss with them, pays for words and a song the cut will not hold.
3 min read
XLSX to CSV in the browser is not a generate
/free-tools/xlsx-to-csv never spends a credit. Do not pay a model to do an in-browser file job.
4 min read