ByteDance Upscaler Video is ByteDance's video/image enhancement model on Versely. This page is its structured prompting reference: the 2 parameters its schema actually exposes, the utility-enhance technique that applies to it, copy-ready templates.
Everything here is grounded in the same sources Versely's agent reads — the model's input schema. Where a line is general craft advice rather than a documented fact about ByteDance Upscaler Video, the page says so.
What ByteDance Upscaler Video wants
The exact input surface, from the same schema the Versely agent fetches with get_model_input_schema before every generation.
| Parameter | What it does | Values |
|---|---|---|
video_urlreq | Source video to upscale | string |
target_resolution | Target resolution (no 720p per docs) | 1080p · 2k · 4kdefault: 1080p |
- — Verified against fal.ai/models/fal-ai/bytedance-upscaler/upscale/video
Technique that applies here
Upscale/enhance/background-removal: mostly parameter-driven, little to no prompt
- None of the three models in this family are reachable through Versely's AI-enhance step at all. POST /features/generate-prompt only accepts content_type image, video, or speech — there's no branch for these tools because there's no prompt to enhance in the first place: brio-video-background-removal and bytedance-upscaler-video have no prompt key anywhere in their schema, and the agent's own dispatch tools confirm it end to end — upscale_image and upscale_video take only model and a file URL, remove_background takes only model and video_url/image_url. The "prompt" for this family is which parameter values you pick, not a sentence you write.
- Background removal — Bria's video model, dispatched through the agent's remove_background tool by passing video_url — has background_color as its real creative choice: an 11-value enum (Transparent, Black, White, Gray, Red, Green, Blue, Yellow, Cyan, Magenta, Orange) that defaults to Black, not Transparent. Naming Transparent explicitly is what gets you an alpha-ready clip; naming a specific color instead gets you a clean, consistent backdrop to key against in a downstream editor. preserve_audio defaults to true and is worth flipping off if the clip's audio is being replaced anyway.
- Video upscaling — Bytedance Upscaler Video, dispatched through the agent's upscale_video tool (model plus video_url, no prompt) — has enhancement_preset as its closest thing to a style prompt: match it to where the footage came from (general, ugc, short_series, aigc, old_film) instead of leaving it on the default, and pair it with enhancement_tier (fast / standard / pro) to trade speed for quality. The schema's own notes call out that there's no 720p option in target_resolution (1080p / 2k / 4k only) — this is an upscale-only path, not a general-purpose resizer.
Copy-ready templates
Replace the bracketed slots; each template says when it's the right shape.
video_url: [SOURCE_VIDEO_URL] | background_color: Transparent | output_container_and_codec: [an alpha-capable option, e.g. webm_vp9] | preserve_audio: [true/false]
Use when: You need a subject-only clip with a real alpha channel to composite over other footage — a UGC overlay, a lower-third, a floating product cutout.
video_url: [SOURCE_VIDEO_URL] | background_color: [Black/White/Green/Blue/... — whatever your downstream keyer or compositor expects] | preserve_audio: [true/false]
Use when: You're keying in a new background yourself in an editor and just need a clean, consistent solid backdrop.
How the Versely agent does this automatically
You can use this page by hand, or let the agent apply the same knowledge. Four real mechanisms — no more, no less:
get_model_input_schema— before generating, the agent looks up ByteDance Upscaler Video's exact input fields, required fields, allowed values, defaults, and min/max bounds. The parameter table above is that same surface.- The prompt enhancer's family rules — 12 per-family rewrite rules (this model's family isn't one of the 12, so only general enhancement applies) shape how a rough prompt gets rewritten.
- The per-provider speech guide — for TTS scripts, the agent follows a provider-specific tag scheme — not relevant to this model, but it's why voiceover scripts come out marked up correctly.
expand_movie_scene— in movie flows, brief scene ideas are rewritten into detailed cinematic descriptions before generation.
Mistakes that waste generations
- Writing a text prompt at all. There's no prompt field on any of these three models — describing what you want in a chat message and expecting it to steer the output does nothing here; the only inputs that matter are the file URL and the enum/boolean parameters.
- Asking the Bytedance upscaler for 720p output, or feeding it a video expecting a downscale. target_resolution's enum is 1080p/2k/4k only, and the schema notes explicitly call out that 720p isn't offered — this is an upscale path in one direction, not a general resizer.
- Sending a video into SeedVR Upscale, or an image into the Bytedance video upscaler. SeedVR's mode is image-only — one reference image in, one image out; video upscaling is a completely separate model in this family, and neither one will silently convert the other's input type for you.
Frequently asked questions
Does ByteDance Upscaler Video support negative prompts?+
No — ByteDance Upscaler Video's published schema has no negative_prompt parameter. Exclusions have to be phrased positively inside the main prompt, or dropped.
How does the Versely agent know ByteDance Upscaler Video's parameters?+
Before generating, the agent calls its get_model_input_schema tool, which looks up the exact input fields, required fields, allowed values, defaults, and min/max bounds for the model. Nothing on this page is guessed — it is the same schema surface those tools read.
Related prompting guides
Generate with ByteDance Upscaler Video
ByteDance Upscaler Video is live in Versely — paste a template above, or just describe what you want and let the agent map it onto the schema for you.