The two files
| Source | FlashVSR | |
|---|---|---|
| Resolution | 1344×768 | 2520×1440 |
| Frames | 158 | 158 |
| Frame rate | 24 fps | 24 fps |
| Length | 6.58 s | 6.58 s |
| Audio | AAC stereo | Same track |
flashvsr_v1.1_tiny_long_bf16 Your finished video at 1080p or 1440p. Every frame, the exact frame rate, the full picture, and the original audio are kept.
Runs on the Sogni Supernet. Nothing to install.
FlashVSR v1.1 is Sogni's video upscaler. Give it one finished video, such as a MiniMax H3 render or a clip you upload, and it returns a sharper copy at 1080p or 1440p on the short edge. It is promptless: the output size is the only choice.
The result keeps the source's frame count, its exact frame rate (fractional rates such as 23.976 included), the full aspect ratio, and the original audio track. Each edge is enlarged up to two times, so 1080p needs a source at least 540 pixels on its short edge and 1440p needs at least 720.
The Sogni SDKs for JavaScript and Python (5.37.1 and later) submit upscale jobs. Send the video and the output size. The server probes the upload for its frame count, frame rate, and dimensions, and prices the job from those verified values.
import { FLASHVSR_VIDEO_UPSCALE_MODEL_ID } from '@sogni-ai/sogni-client';
const project = await sogni.projects.create({
type: 'video',
network: 'fast',
modelId: FLASHVSR_VIDEO_UPSCALE_MODEL_ID,
positivePrompt: '',
numberOfMedia: 1,
referenceVideo: fs.readFileSync('./clip.mp4'),
upscaleResolution: 1440 // or 1080: the output's short edge
});
const [upscaledUrl] = await project.waitForCompletion(); // MP4 with the original audiofrom sogni_client import FLASHVSR_VIDEO_UPSCALE_MODEL_ID
project = await sogni.projects.create(
type="video",
network="fast",
model_id=FLASHVSR_VIDEO_UPSCALE_MODEL_ID,
positive_prompt="",
number_of_media=1,
reference_video="clip.mp4",
upscale_resolution=1440, # or 1080: the output's short edge
)
print(await project.wait_for_completion()) # MP4 with the original audioFrame count, frame rate, width, and height are optional; any you send must match the source. The SDK README shows how to request a price estimate first.
The comparison at the top is the clip from our upscaling comparison: a 6.6-second MiniMax H3 video at 1344×768, upscaled by FlashVSR v1.1 to 2520×1440. Both files keep all 158 frames at 24 fps and the same soundtrack.
The two files
| Source | FlashVSR | |
|---|---|---|
| Resolution | 1344×768 | 2520×1440 |
| Frames | 158 | 158 |
| Frame rate | 24 fps | 24 fps |
| Length | 6.58 s | 6.58 s |
| Audio | AAC stereo | Same track |
Time for this clip
| Upscaler | GPU | Total time |
|---|---|---|
| FlashVSR v1.1 | RTX 5090, tuned | 1 min 45 s |
| FlashVSR v1.1 | RTX 4090 | 3 min 6 s |
| SeedVR2 7B | RTX 4090 | 10 min 48 s |
Measured in the comparison on this clip, including model loading and video export. The 4090 times are one run each; the 5090 time is the median of three. On the Supernet, the worker that takes your job sets the time.
FlashVSR v1.1 delivered the most overall detail enhancement and our favorite quality/speed balance. Sogni Labs, The Best Open-Weight 2K Upscaler for MiniMax H3 Video?
Upscale the cut you approved: a MiniMax H3 render, another model's output, or a clip you upload.
The result has the source's frame count, so timing, motion, and cuts stay where you made them.
24, 25, and 30 fps, and fractional rates such as 23.976 and 29.97, come back unchanged.
Dialogue, music, and effects are carried over from the source file.
Portrait or landscape, the full frame is enlarged at its original aspect ratio.
NVIDIA GPUs on the Sogni Supernet do the work. No local GPU or setup required.
FlashVSR is a standard Supernet model. A Sogni Unlimited subscription covers your upscales under fair use, and Premium Spark covers them without a subscription.
Sogni Unlimited
$20 / moFair-use Supernet rendering with no credit counting. Video upscales are included with the rest of the Supernet catalog.
Compare plans →Premium Spark
from $1.99Pay as you go on the fastest priority queue, with no plan to manage.
Buy Spark PacksAn upscale is priced on the source's length and size. The 6.6-second reference clip estimates at about 53 Spark on the Fast network, and 1080p costs the same as 1440p because the model does the same work for both. Sogni shows the estimate before the job runs.
A source up to 768 pixels on its short edge and 100 MB, 1 to 362 frames (about 15 seconds), at a constant frame rate from 1 to 60 fps. It should be SDR, with square pixels and any rotation already applied.
It sharpens detail and keeps the content: the same frames, frame rate, framing, and soundtrack. It is promptless, so there is nothing to restyle.
Choose the largest size your source allows. 1440p (2K) needs a source at least 720 pixels on its short edge, and 1080p needs 540. Both cost the same for a given source.
It is covered by a Sogni Unlimited subscription ($20/mo or $199/yr; Unlimited Pro is $50/mo or $498/yr) under fair use, with no credit counting. Without a subscription it runs on pay-as-you-go Premium Spark, and Spark Packs start at $1.99. The 6.6-second reference clip estimates at about 53 Spark on the Fast network. See the pricing breakdown ↑
In our comparison, the 6.6-second reference clip took 3 minutes 6 seconds on an RTX 4090 and a median of 1 minute 45 seconds on a tuned RTX 5090, including model loading and video export. On the Supernet, the worker that takes your job sets the time.
Sogni offers 1080p and 1440p on the short edge, up to 2560×1440 pixels. The 1344×768 reference clip becomes 2520×1440 at 1440p.
The linked Sogni Labs article compares FlashVSR with SeedVR2, NVIDIA RTX VSR, FFmpeg Lanczos, and a community latent upscaler on matched clips, with synchronized playback and 100% detail views.
Upscale a finished video in Sogni, or submit it from the SDK.