Edit with music and captions in one pass

One edit_video call can stitch clips, mix a bed, and burn captions. Edit a video with music and captions is that assembly. Preview at 480p. The final export is one editor_export whether you stacked music, type, and a ratio or not. Do not chain cut_video + attach_audio_to_video + add_veed_captions for a three-clip UGC unless you needed those as separate reusable files.

This is one pass on keepers you already have. It is not “mix a bed under native audio” as a standalone radio, and it is not captions-after-generate without the rest of the ad.

One call is the assembly

edit_video’s own description prefers this over chaining cut + attach + captions when the user wants a multi-clip assembly in a single step. The only required field is clips: HTTPS URLs, ordered. Everything else is optional and in the same call:

• transition / transition_duration between clips. Hard cut is the UGC default. A long crossfade is a brand film.
• texts for authored on-screen copy.
• music and voiceover for audio.
• overlay for a talking-head PiP.
• captions to burn speech.
• aspect_ratio and resolution for the export window.

Clips must be HTTPS. file:// is rejected. Returns { video_url, credits_charged, preview_notes } or a 402-style error if the balance cannot cover the final. check_credits before a big job. estimate_cost if you need the number.

What “one pass” is for:

• Three UGC keepers (hook, demo, CTA) + instrumental + captions + 9:16.
• A talking-head plus a pack insert, one bed, one burn.
• A mixed-ratio mess you are standardizing while you finish.

What it is not for:

• Hunting the take. Generate first. Assemble keepers.
• A 20-clip documentary you have not watched. Preview will not save an unwatched EDL.
• Replacing a baked native-audio line and keeping it. Say mix or replace. If the keepers include Veo 3.1 (40 credits, native audio) and a silent Turbo clip, the seam will pop unless you name audio handling.

Suno Sounds V5.5 is 2 credits if you still need to write a bed. Generate the instrumental first if you do not already have a URL. Prompt without a vocal. A singer under a talker is two mouths; captions will hear both.

Preview 480p until the mix and the burn both read

preview:true is a free 480p render, per-user cooldown. The default preview:false is the paid export. Iterate on preview. Do not pay to discover the bed is louder than the line.

Preview checks:

1. Order. Hook, proof, CTA. Wrong order is a different ad.
2. Seams. Three frames either side. Hard cuts on UGC. If a seam needs B-roll, you listed the wrong clips, not the wrong dissolve.
3. Bed. Under the voice, not over it. Duck. If a keeper already has native-audio score, a second bed is mud. Mute the generate’s music only if you are willing to lose baked speech too — native audio is not stems. Silent keepers plus one VO plus one instrumental is the clean stack.
4. Captions. First line readable muted, off chrome. Safe area on 9:16 is username, caption field, right rail; Shorts may add a photoreal AI overlay on the picture (YouTube Blog, 27 May 2026).
5. Ratio. 9:16 if that is the ship window. Do not preview 16:9 and export 9:16 without looking.
6. PiP. If you used overlay, it must miss the burn and the rail.

Caption facts if the pass burns speech: add_veed_captions family — 165 language codes, 21 BASIC at 1×, 9 DYNAMIC at 2×. DYNAMIC as part of a one-pass export still bills at 2× for that look. Preview the style on this background. Glass over a white pack is a different problem than glass over a dark room.

Authored text in the same call is not a transcript. A URL nobody said belongs in texts, not in captions. Mixing them is how karaoke highlights a price the speaker never read.

If preview is wrong, change the instruction and preview again. The cooldown exists so this is not an infinite free renderer. Use it for real questions (is the bed legal, is the crop clipping the cap), not for twenty fonts.

Do not chain three tools for one ad

Chaining is correct when you need a reusable intermediate:

• A trimmed clip you will also post unsliced.
• A VO wav you will attach to three ratios.
• A caption style you are still picking with preview_caption_style on a single file.

Chaining is waste when the only output is one TikTok. Each tool is a wait and a chance to caption the wrong parent. The one-pass door exists so the EDL is one object.

Limits:

• Natural-language EDL, not a frame-by-frame speed curve editor. Fine ramps belong in preview.
• One export price for the final regardless of how many features you stacked. That is the point. It is not a reason to stack features you do not need.
• 402 on insufficient credits rather than a partial charge. Check the balance.

Add captions to a video remains the right door for a single file that is already picture-locked and only needs a burn. Add music to a video remains the right door for mix-under-native-audio on a finished picture. This page is when those jobs arrive together with a stitch.

Do not one-pass a file you have not picture-locked. A later swap of the middle clip means a new assembly. Keep the parents.

What still has to be true before you assemble

Keepers, not maybes.

• Each clip already does its job. Merge of maybes is a longer maybe with captions.
• Ratio intent named. Crop stills before they were I2V’d if you can. The assembly will reframe; it will not invent a vertical set.
• Script locked if you are attaching VO. Duration of the stitch is the budget for the read.
• Brand look named: BASIC vs DYNAMIC, hex if you have a kit. Do not say “make it branded” and accept a cousin preset.
• Disclosure plan. Article 50 (2 August 2026) at first exposure for photoreal AI. YouTube long-form label under the player; Shorts overlay on the file. The one-pass export does not add those for you.

The UGC video generator produces clips. This tool finishes a set of them. Generate is not a studio. Assembly is.

Run the one-pass edit in Versely: clips, bed, captions, 9:16, preview, then export. Do not pay three renders to build one ad.

FAQ

Why not chain cut, attach, and captions?

You can when you need the intermediates. For a multi-clip assembly the editor prefers one edit_video call: stitch, music, voiceover, overlay, captions, ratio together.

Can I preview before it costs credits?

Yes. preview:true is a free 480p render with a per-user cooldown. Final export is the charged render.

Does stacking music and captions cost extra on the export?

The final is the standard editor_export price for that render, not a la carte per feature. DYNAMIC caption looks still sit on the 2× caption tier when that look is what you burned — ask estimate_cost if you need the number for your file.