PRACTICAL GUIDE · PROMPTTOVISUAL
Vidu Q4 and Q4 Preview: Modes, API Limits and Channel Pricing
Vidu Q4 makes video from a starting image or multiple references, but Q4 branding and Preview API contracts differ by channel. Compare exact inputs, audio controls and published charges before choosing. PromptToVisual has not integrated Vidu.
EXTERNAL MODEL GUIDE · CHECKED OCTOBER 9, 2026
Explore Vidu on the official site → · Read the official Preview API contract
No Vidu generation is available in PromptToVisual. Start with the exact channel and mode below.
Start with the mode, then choose a channel
Use image-to-video when one image should anchor the opening frame. Use reference-to-video when several subjects or a voice reference need to guide the scene. Vidu advertises synchronized audio, multi-shot sequences and improved motion; those are vendor claims, not our quality or speed measurements. Its product page uses Q4 while its API documentation still names viduq4-preview. A shared family name does not prove identical deployed weights or behavior across services.
Official Vidu Preview: a single starting frame
The official POST /ent/v2/img2video contract accepts model viduq4-preview and one image through images. It documents integer durations from 3 to 16 seconds, 540p, 720p, 1080p, 2K or 4K, and an audio boolean for sound versus silence. Defaults are five seconds, 720p and audio on. PNG, JPEG/JPG and WebP are listed, with a 50 MB image limit but a separate 20 MB HTTP body limit. Account access, content acceptance and final output have not been tested here.
Official Vidu Preview: multiple subjects and voices
POST /ent/v2/reference2video uses viduq4-preview, a prompt and 1–15 images. Optional sounds supports up to three MP3 references, each 3–12 seconds and up to 50 MB, subject to the separate request-body limit. Duration is 3–16 seconds; resolution choices run from 540p to 4K. The output audio switch is separate from providing voice references. Five aspect ratios are documented: 1:1, 9:16, 16:9, 3:4 and 4:3. Only use images and voices you have permission to provide.
fal: image-to-video has its own input shape
fal-ai/vidu/q4/image-to-video requires image_url rather than Vidu’s images field. The checked schema lists 3–16 seconds, five resolutions from 540p to 4K, and a prompt limit of 5,000 characters. The page describes native audio, but this image endpoint’s checked input schema does not expose the official API’s audio boolean. Do not send another channel’s optional fields or assume a silent-output switch exists. Exact current fal pricing was not verified from the retrieved public page; check the endpoint quote before running.
fal: reference limits differ from the official platform
fal-ai/vidu/q4/reference-to-video documents up to 12 reference images and three MP3 voice clips, not the official platform’s 15-image maximum. It exposes reference_image_urls, reference_audio_urls, an audio switch and five aspect ratios; duration and resolution choices are 3–16 seconds and 540p through 4K. Its route says q4 while descriptive text still says Q4 Preview. Exact equivalence to a non-Preview release is UNKNOWN. Select by this endpoint’s current schema, and keep safety screening enabled.
WaveSpeed image-to-video: settings and a bounded quote
For vidu/q4-preview/image-to-video, provide one image URL. The page lists integer 3–16-second durations and generate_audio plus enable_prompt_expansion, both defaulting on. Its published USD rates per second are 540p $0.045, 720p $0.095, 1080p $0.12, 2K $0.19 and 4K $0.39. It states only duration and resolution change this quote, so five seconds at 720p is $0.475 with the default audio setting. Checked October 9, 2026; no promotion end date is stated. This is a channel quote, not a universal Vidu rate or tested bill.
WaveSpeed reference-to-video: read the conflicting image limit
vidu/q4-preview/reference-to-video lists prompt, optional images and audios, and generate_audio. Its prose says up to 12 images, while the visible form counter shows 15; the accepted maximum remains unresolved here. The page lists the same five USD per-second rates: $0.045/$0.095/$0.12/$0.19/$0.39 for 540p/720p/1080p/2K/4K. It states other supported settings add no separate charge. At default audio on, 720p and five seconds, the published quote is $0.475. Checked October 9, 2026; no promotion expiry stated. Confirm the limit and account quote before use.
Official pricing uses a different credit unit
Vidu’s pricing table lists viduq4-preview image-to-video and reference-to-video at 9/19/24/38/78 Vidu credits per second for 540p/720p/1080p/2K/4K. The displayed credit unit is RMB 0.03125. A five-second 720p calculation is 95 Vidu credits, or RMB 2.96875 at that unit. The Q4 table does not separately price audio on/off; do not apply another model’s audio surcharge or off-peak discount. Checked October 9, 2026; no Q4 promotion expiry identified. The page asks for account information to view full pricing, so verify your final account terms. Vidu credits are not PromptToVisual credits.
How to make the choice without a misleading lowest price
For the official workflow, start at Vidu and check the exact Preview API contract if automating. For fal or WaveSpeed, choose the endpoint matching your inputs, then confirm duration, resolution, audio, reference limits and the account-specific charge together. A useful comparison records one identical requested output configuration per channel; do not compare a minimum-resolution quote with a 4K sound-enabled task. No new sample or independent benchmark was produced for this guide. Public downloadable weights and a license for self-hosting were not verified; this does not mean weights can never be released.
What is available in PromptToVisual?
Vidu is not integrated here. PromptToVisual’s separate Seedance 2.5 workflow uses one permitted first frame plus a prompt for a four-second video with audio, square or landscape, at 40 paid credits. It requires verified sign-in, subscription or add-on credits rather than trial credits, and runtime and safety eligibility. This is a different model and narrower workflow, not a substitute contract for Vidu’s multiple references, voices or longer sequences.
Sources and next steps
Sources checked October 9, 2026. Each channel and Preview endpoint has its own contract. Published quotes are not tested bills; no integration or independent model test.
Official Preview reference API →
WaveSpeed Preview image contract and quote →
WaveSpeed Preview reference contract and quote →
Utopai X access guide · Keyframe control guide · FLUX 3 Video guide · Seedance limits · PTV single-first-frame alternative · All guides