Video to Video on ToVideo
Upload a reference clip, style frames, or audio — then prompt the change you want. Follow motion, lock a look, or extend an edit without inventing everything from text alone.
Start from a source you trust
When video to video beats text or image alone
Choose Reference (video to video) when an existing clip, style set, or audio bed should steer the output. Text invents without uploads. Image animates a still you already locked. Reference follows sources you attach — video, images, or audio depending on the model. The patterns below are the jobs teams run once they have something to follow.
Motion follow from a short clip
You have a 2–15s take with camera path or body action you like. Prompt the new subject or setting while asking the model to keep timing and movement energy. Keep the reference short and clean — busy multi-scene masters confuse guidance more than they help.
Style lock with reference images
Some models accept image packs as look references (palette, costume, rendering). Upload 1–N frames that already show the brand look, then prompt the action. Prefer consistent lighting across refs; mixed art styles in one batch usually muddy the result.
Audio-led rhythm or dialogue bed
When a model supports audio references, attach a short WAV/MP3 bed so lip motion, cuts, or energy can track the track. Keep duration within the model’s limits. Do not expect perfect lip-sync marketing claims — treat audio as guidance, then cut in an NLE if needed.
Edit, extend, or restyle a take
Certain models treat the first video as the edit or extend target and extra clips as supporting refs. Use that when you need a continuation, a restyle pass, or a controlled rewrite of an existing MP4 — not when you still lack any source footage.
How to brief Reference without fighting your sources
Video to video fails when the prompt contradicts the upload or when you stack too many conflicting refs. Treat sources as the motion/style law; use text for the delta. Recipes below are written for ToVideo operators.
What ToVideo adds for video to video
Differentiation is workflow, not another slogan: Reference mode with model-aware slots, visible credits, and the same account as Text and Image.
Reference mode, model-aware inputs
Open the video generator and choose Reference. Image, video, and audio slots appear only when the selected model supports them — including limits on count, format, and length.
Credits before Generate
Cost updates with model, duration, quality, audio, and whether video references are attached. Failed finals refund so you can refine guidance without silent balance burn.
Text vs Image vs Reference
Text invents with no upload. Image animates stills. Reference follows clips, style frames, or audio. Switch modes in one form; unavailable options stay hidden.
Same account as stills and text runs
One login and credit pool across AI Image and Video. Draft in Text, lock a still in Image, or guide with Reference — without exporting between unrelated tools.
Server jobs + AI Tasks
Close the tab; the run continues. Download the MP4 from the preview panel or AI Tasks when status succeeds.
Download-ready MP4
Successful generations expose download in the result area. Archive outside ToVideo for ads, clients, or your editor.
FAQ
Operator answers for reference-guided video to video.
Ready to guide with a reference?
Attach one clear source, prompt the delta, check credits, and generate.